High Voltage Probe Design Objectives, Principles and Applications
Written by R. J. Adler, North Star High Voltage Corporation
Electricity has powered advances – literally and figuratively – over the last 150 years leading to the most prosperous human civilisation which has ever existed. Measurement (in particular electrical measurement) has a significant role in this progress. We can only improve equipment and processes if we know quantitatively and in detail what power/voltage/current etc. are produced. That is the role of measurement.
The Egyptians recognised the importance of measurement and had a goddess of writing and measurement. By 3100 BC they had created standards of length which they stored and referred to. Figure 1 shows a scale used to measure the height of the Nile and particularly the height of the flood.
Figure 1: Egyptian “nilometer” for taking the most important measurement in Egypt – the height of the Nile. This device dates to 1500 BC or earlier.
In our modern electrically actuated and electrically recorded society, measurement of voltage and current are among the fundamental measurements which allow us to understand technologies of all types. High voltage measurement – which for no particular reason we define as voltages above 1 kV – has a major role in measurement science and technology development. A quantitative reason for choosing 1 kV is that it is a voltage significantly above the Paschen minimum voltage of about 350-400 V at STP, so air breakdown is primarily a gas effect.
North Star High Voltage (NSHV) has developed an easy to use, high voltage measurement solution, with a range of precision, wide bandwidth probes. They are designed for use with oscilloscopes and meters with >1 MΩ input impedances and are suitable for DC, 50 Hz AC, or pulsed measurements.
2.0 Application Areas
More information about the wide variety of technologies and applications the probes have been used for can be found here.
The requirement for the many applications is of course perfectly precise voltage measurement at all frequencies, so that must be our objective. Inevitably there will be some compromise.
3.0 RC Resistive/Capacitive Voltage Dividers
A resistive divider can readily measure DC signals in conjunction with an oscilloscope or meter. A capacitive divider with sufficient capacitance can readily measure AC signals. For Zhv consisting of Chv and Rhv, as well as Zlv consisting of Clv and Rlv we can readily calculate the complex attenuation ratio.
A = 1+ Zhv/Zlv
= 1+ (Rh/Rl)* [(1+ω2τhτl+iω(τl-τh)]/( 1+ω2τh2)
Where τh = RhvChv and τl = RlvClv
Click here to expand
Figure 2 Schematic of an idealised RC mixed capacitive divider
The obvious preferred case is where τl = τh, A = 1+ (Rhv/Rlv) = 1+ (Clv/Chv) in which case the signal output is attenuated and the input and output are perfectly in phase. Note that in this idealised case, the attenuation is smoothly identical from DC to the highest frequencies without any anomaly at ω = τ. In some cases, and over limited parameter ranges, such a divider can be built without compromise but the problems in the next section need to be considered.
Click here to expand
Figure 3 Some of the stray capacitance effects in two general types of RC high voltage probes
3.1 More Complete Model of a Probe
In Figure 3 we show more stray capacitive coupling, which affects the response of the frequency probe. A typical step function response is shown in Figure 4. The curvature is due to the stray capacitances of Figure 3. The curvature shown is relatively small as curvatures (inaccuracies) of 10 – 20% with typical geometries are not uncommon.
Click here to expand
Figure 4 Step function modified by stray capacitance effects
An extreme example of the capacitive coupling effects is typical, low priced, meter DC probes. Their inaccuracy at 50/60 Hz is typically 10% due to their high resistance and capacitive coupling.
It’s important to understand that the frequencies at which the capacitive coupling occurs are quite broad. For example, if there is a 1 pF capacitive coupling to a 100 MΩ of resistance, the nominal time constant is 100 μs and the nominal frequency is 1.6 kHz. Lowering the resistance increases the frequency of maximum deviation from ideal behaviour. In rough terms at 100 kΩ, the nominal frequency of maximum capacitive deviation is 1.6 MHz; still within the band of interest for many measurements. This means that simply trying to use very low resistances is not an effective strategy for avoiding the effects of capacitive coupling.
It should also go without saying that reducing the resistor value quickly increases the probe power dissipation. A 100 MΩ resistor at 40 kVDC dissipates 16 W. We generally keep the resistance at 100 – 200 MΩ per 10 kV (0.5-1 W/10 kV).
Increasing the spacing between conductors is a very slow way to decrease capacitance – particularly when we note that a sphere has capacitance to infinity of value 1.1 ɛpF*R (cm) in vacuum. Meaning that a sphere of 2 cm diameter has a minimum capacitance of 1.1 pF – although of course the grounded (or high voltage) surface to which it couples is usually closer than infinity and therefore more capacitive. Reducing capacitance in other ways by adjusting geometry is possible, but in general reducing a radius increases an electric field, thereby increasing the high voltage insulation problem.
We have instead developed techniques to compensate accurately for stray capacitance in order to produce high fidelity outputs. The RC probe response depicted in Figure 4 was corrected using our techniques to produce the output of Figure 5. Typically we correct the response to 0.5% to allow for other effects within the 1 – 1.5% typical AC specification. Our capability to accurately correct the capacitive effects is the reason that we can use oil insulation in the high voltage section without significant stray capacitive effects. The use of oil helps to deliver a very low (almost zero) rate of high voltage failures.
Click here to expand
Figure 5 The device of Figure 4 corrected for the effects of stray capacitance
4.0 DC Measurements with the Probe or Divider
NSHV probes measure DC accurately – to 0.1 or 0.2% in most cases. The important issues in making such a measurement are resistor linearity, control of electric field configurations and control of extraneous (unexpected) currents. A typical resistor will vary in its resistance by 0.3% or so over a typical 300 V range of use. The situation is much worse for high voltage resistors without careful attention to resistor design and selection. With careful selection, resistors with typical voltage variations of 0.1% over their range of use are found in the NSHV probes.
High voltage resistors have highly convoluted paths in order to reduce electric fields by increasing the conduction length for a fixed voltage. NSHV uses special resistors made exclusively for use in the probes. They are tested on a batch basis, and to full voltage in many of the probes and dividers. A measured attenuation ratio of an optimised NSHV VD-500 (which had hand-selected resistors) as measured by a European National Standards laboratory is shown in Figure 6.
Click here to expand
Figure 6 Measured attenuation ratio of an optimised North Star VD-500
In the case of the VD-500, the maximum deviation from the calibrated value is 0.05%. When we prepare the 0.1% uncertainty specification, it is the allowable deviation from the stated attenuation ratio in a typical laboratory situation (i.e. your application). For example, 0.1% uncertainty means that for standard laboratory temperatures (15-30 °C) and including variations in linearity, the divider will be within 0.1% of the stated value. If the ratio is stated as 10,000:1 at 0.1% (standard ‘VD’ probe specification) the actual ratio at all voltages and temperatures will be between 9990:1 and 10,010:1. The precision and reproducibility are much better than the 0.1% value claimed. Observations from customer calibrations performed at NSHV is that the DC values and linearity remain fixed for up to 20 years in 80% of cases.
The 0.1% accuracy value is not a noise value because the intrinsic noise of the dividers is much lower than 0.1%. Conventional concepts of resistor noise are modified by the relatively large amount of parallel capacitance in NSHV dividers which act as a low pass filter for noise (in other words the divider attenuates frequencies typically associated with resistor noise).
A standard production PVM-100 DC attenuation curve is shown below (note the DC spec is 0.15% for the PVM-100). PVM-1, 2, 3, 4, 5, 6, 7 and PVM-100 are typically like this.
PVM-100 S/N M22420
|Vset(Agilent)||Vin Meas||Tap Voltage||Voltage Ratio|
5.0 DC Insulation
NSHV has implemented a very conservative electric field design for the probes. For example, the DC insulator length between electrodes is 35 cm for a VD-100 and the voltage rating is 100 kVDC. The gradient is 2.9 kV/cm DC and 4.8 kV/cm pulsed. Breakdown along the insulators is exceedingly rare and has never been reported for a NSHV probe used within these ratings.
Note that many “standards” (ISO Standards etc.) quote much lower values and allow convolution of the surface to achieve longer creepage lengths. These concepts originated in situations where there are very large peak fields (such as the edge of a circuit board trace) or where there is rain (leading to the convolutions seen on outdoor insulators). The extension of these ideas to high voltage probes exemplifies science and engineering created by committees.
Because NSHV probes have a very low insulator gradient, electrodes (termed corona rings or toroids in the lexicon) can be added to the structure in order to reduce stray capacitive coupling between the central column (with resistors and capacitors fitted inside) and ground. A pulsed mode electric field plot of the VD-100 is shown in Figure 7.
Figure 7 is a cut through of half a probe, laying on its side. At full rated peak voltage (160 kV) the max field is on the bottom surface of the top (right in Fig 6) electrode at 29.8 kV/cm (max of 18 kV/cm DC). This field is non-uniform and so the appropriate corona inception field is roughly 40 kV/cm. Therefore the probe is generally corona free up to the full pulsed voltage. The blue lines are lines of electric field and one of the key points is that the top field couples to ground over a significant fraction of the corona shield. This is asymmetric and requires correction beyond the simple RC ratio method. North Star is able to correct the response of the probe to 0.4% typically – well within the typical 2% AC specification.
Click here to expand
Figure 7 Field plot of a VD-100 probe with its axis horizontal and ground vertically. The axial centre is on the bottom.
6.0 Proximity Effect (Stray Capacitance) of the Probe
The AC response of the probe varies with proximity to a grounded surface, such as a metal wall. Many walls have wires in them (most in fact) and these wires make most walls into approximate ground planes. In Figure 8 we show one of the larger probe’s proximity effects – in this case a VD-150. Despite the fact that this is larger than others (larger than the VD-60, VD-100, VD-200, for example) at only 7” or 177 mm from the edge of the probe, the proximity effect drops to 1%. The 20 and 50 kHz are shown for illustration – there is no significant difference between the two. The nominal transition between the proximity and no proximity frequency is at about 100 Hz. The proximity effect at 50/60 Hz is about half that shown in Figure 8.
A VD-150 probe with a ground at only 7” is problematic from the HV breakdown standpoint so proximity is not a practical problem with NSHV VD probes up to 200 kV. There is larger proximity effect in the VD-300 and VD-400.
Click here to expand
Figure 8 Effect of proximity on ratio for a VD-150. VD-100 and VD-60 probes have lower proximity effects.
Proximity Effect – PVM Probes
PVM 1, 2, 3, 4, 5, 6, 7 and to a great extent PVM-100 are shielded from proximity effects by the outer cylindrical grounded conductor fitted to the probe. North Star can define the maximum proximity effect as “the proximity effect when the probe is resting on the cylindrical conductor and rested on a ground plane”. The minimum proximity effect occurs when the probe is in free space far from grounded conductors (but perhaps not infinite).
For the PVM1, 2, 3 the proximity effect leads to a change in the ratio of 0.2% or less. At that value it’s difficult to quantify. The PVM-5 and 6 have proximity effects of about 0.4%. The PVM-4 and PVM-7 probe proximity effects are about 2x higher at 0.4% and 0.8% respectively. The PVM-100 has higher proximity effect in the range of 1-2%.
Users have described higher proximity effects in extreme cases such as when the probe body goes through a close fitting vacuum feedthrough or when the nose of the probe is immersed in oil. For unusual situations like these, NSHV can test proximity at any high frequency or with a fast transition. This correction will apply for all frequencies above 1/2 πRC where RC is given in the calibration report provided with the probe. This is in general the frequency dividing line between AC and DC behaviour.
The PVM-12 has a higher proximity effect, because this probe doesn’t have the larger outer cylindrical conductor. The proximity effect is less than 2% if the ground is kept away from the probe body by the distance defined by the probe rest (plastic disc). When proximity effects are observed with the PVM-12, it’s usually when conductors are very close to the probe body.
7.0 Measuring Devices (Oscilloscopes & Meters)
North Star devices are all designed for use with 1 MΩ input oscilloscopes. For reasons lost in history, oscilloscopes have an input impedance of (almost always) 1 MΩ, and occasionally 50 Ω as a switchable option. Meters tend to have 10 MΩ or infinite input impedance (with a smattering of other values used by various vendors). The probes can also easily be used with 10 MΩ input impedance meters, by adding parallel resistors to make them “look” like 1 MΩ.
The method of correcting the probe output to “read” correctly is shown in Figure 9. NSHV offers two methods of making this simple transition. The first is a parallel BNC to Banana plug arrangement. The second is a built-in switch, with a switch between two parallel resistances. These are shown in Figure 9.
Click here to expand
Figure 9 Standard ways of using a North Star probe or divider. In the top picture, plug the probe into a scope with the provided BNC cable. Alternately use a BNC to banana adapter (middle) with an impedance match resistor (1.111 Meg) and the switch option (low).
8.0 Elimination of Transmission Line Reflections
NSHV probes have usable bandwidths up to 150 – 200 MHz and use standard coaxial cables, yet cable reflections are negligible (you would expect a transmission line with 50 Ω or 90 Ω characteristic impedance would see reflections when terminated by a high impedance load). Other probes might use resistive cables which tend to limit the length of cables and flexibility of cable length. By using standard coaxial cables without external terminations or reflections, North Star can produce probes with cable lengths of 1 m – 30 m (or more).
Making accurate and wide band measurements of high voltage signals is challenging and there are many trade-offs and considerations. NSHV has studied these issues to deliver a high voltage measurement solution that is easy to use.