In this blog, I want to share a very nice experience that came out after needing a sensor to capture a very fast and short current pulse at the HV lab.
A high-frequency current transformer most likely pops up in everyone’s mind right away and certainly, there are plenty of them somewhere in the lab. Somehow, it can be said that the gain ( I would learn quickly from manufacturer data sheets that the gain is also referred to as transfer impedance, sensitivity or the figure mV/mA) and bandwidth are the two main specs from an HFCT to check before a quick test in most of the applications. In our case, the size was also an important spec.
Why? Simply because the task was to capture signals flowing along a GIS compartment. As strange as it seems, we needed to slide an HFCT through one of the bolts at the GIS spacer. Difficult to explain to a manufacturer so we gave a try to build our own HFCT.
Complexity aside, an HFCT is a 1:N turns transformer with a ferrite core having N turns wound onto it. Electronics retailers offer many options for magnetic cores at relatively low prices. For example, we chose an EPCOS ferrite core made of N30 material with dimensions 34 x 20.5 x 12.5 mm for around 5 eur/pc from Farnell.
The next step was to wind the turns onto the core, for which some basic rule-of-thumb concepts are worth to have in mind:
- The higher the number of turns the lower the gain and the lower cut-off frequency shifts toward lower frequencies. A good reference for more insights about HFCT design can be found here.
- Many turns will limit the upper cut-off frequency simply because at very high frequencies the stray capacitance between turns will offer a lower impedance path for the current.
- Copper tape is preferred over wire as the conductor for the winding. In our case, we used adhesive copper strips of around 3 mm. Two advantages of this are: the tape is very thin so it contributes to reduce the stray capacitance and being adhesive makes it easy to evenly distribute the turn in the core.
These are some of our attempts.
Once the transformer itself was ready, our technician Wim prepared the housing for mechanical support and also to serve as electromagnetic shielding.
The result was this sturdy solution with dimensions that make it fit into the GIS in the picture back on top of this blog.
If you look closely at the BNC connectors from the left picture, you notice that one BNC is isolated from the housing. This is the output of the HFCT. The next BNC is not isolated and if you connect a cap, this results in grounding the sheath of the coaxial cable. Good to have the option of a floating output or a grounded one.
But before being ready to use our HFCT, we tested its frequency response. A network analyzer is all that is needed for this. However, we want to show you all some sort of a “hack” in case you don’t have one on hand.
The task is to measure the output voltage due to the input current, i.e. the ratio mV/mA. For this, we need an oscilloscope and function generator. We used the Tektronix AFG3252C and the DPO3034. However, any good quality oscilloscope and function generator may serve the purpose, and they also are easy to find in a lab, which is very convenient.
The measuring principle is as straightforward as using the generator to drive a current through the HFCT and measure its output voltage.
This set-up looks easy and actually it is. Perhaps the current loop is what you need to care about.
To assure a neat arrangement we suggest to follow these steps:
- Pass on a short piece of wire through the HFCT and solder BNC connectors to both ends.
- Press the base of the BNC connectors onto the HFCT case. We used a clamp.
- Connect a 50 Ω load to one end.
- Connect the other end to the function generator.
Now, note that the input current is the generator voltage Vgen/50 Ω. Also note that the return path of the current is comprised of the cases of all the elements: HFCT, BNC connectors and the resistor load. The main conductor of the coaxial cable was kept as short as possible and the coaxial structure is broken only this short distance which is good for shielding.
- Set the function generator output voltage to its maximum, Vgen (peak value).
- Sweep the frequency of the sinusoidal waveform. We swept from 10 kHz to 200 MHz which is the maximum frequency of the function generator.
- Measure the peak output voltage of the HFCT, Vout.
The frequency response will result of computing Z = Vout / (Vgen/50 Ω).
As shown in the figure below, the lower cut-off frequency was 62kHz, the upper one was 136MHz and the gain 8.5.
Looking at the figure, I would say that we got a quite nice smooth frequency plot. For lower frequencies certainly this is not a surprise, but we were eager to see how high in frequency we could get with this simple set-up.
At the end of the day, we saved a few thousand euros by crafting our sensor!!!