Current transformer calcs and frequency effect?

[Al (astral highway)]

Hey folks ,

I’m still building, testing and refining current transformers for my project. I say ‘RF’ up there and to be more precise, the design frequency is about 230Khz.

The latest iteration is wound on an Epcos ferrite toroid, on N30 material , Al 8700nH. devoce Under test is a CT I wound with 60 turns of 1.3mm dis Litz wire.

The cores will be the main sensors for feedback and overcurrent. Pulses will be high tens to >hundred A.

My dilemma is that I don’t have a signal source this powerful, so I’m testing and calibrating with an improvised method.

I have passed two turns around the CT so it’s now temporarily a 1:30 device under test (DUT).

A calibrated DC power supply tells me the current and voltage under load of a square wave source with class B follower. Assuming 65% efficiency, I can thus calculate the effective power through a 3R3 resistive load. The current passes untouched the two turn overwond of my CT.

Here are the two ‘tests’.

1) 18Vpp, CT open circuit (no burden resistor ).
DC current drawn from signal source and follower, 10mA. As expected, high voltage across open circuit CT, outside ‘scope 10 X range. Small but perceptible spark if S/C. Current in secondary 10mA/30= just 0.33mA

2) Burden resistor 1R, 5%
18Vpp sine wave at 230KHz

Measured volts ‘(scope) 145mV.
Rise time 77nS.

Effective current into CT approx 200mA. (This is allowing for small but perceptible losses as heat dissipated in the class B follower transistors, assuming 65% efficiency)

I am familiar with the formula for turns ratio and primary and secondary voltage.
Also:

Eo =Ip * Rb/Tr

Where Eo is volts across CT with burden
Tr is Turns Ratio
Rb is Burden resistor

However, these calcs make a nonsense of my result of 145mVpp across the burden resistor (let’s call it 100mV RMS).

So there must be a frequency-dependent variable. The CT is not resonant at the test frequency, but it heads towards resonance if I sweep down to 80KHz. However, there is noteable ringing with a quasi resonant peak and exponential decay envelope, which will skew the measured output voltage .

Please note, once I’m clear about the test protocol, the CT output will be full wave rectified and a DC voltage will be allowed to develop in a suitable capacitor before a comparator stage. But I want to be clear about its characteristics before I complicate things with rectification of the output.

Long story short, what’s the fool-proof way to evaluate my CT without having to use more complex equipment and what’s the deal with frequency in my calibration calcs?

Over to you, magnetics experts !!

P.S - another thought is, what attenuation arrangements would result in my 1:60 device, with this documented response , having a sensible primary current to secondary voltage ratio.

Thank you

[Argus25]

Al,

To be 100% sure you will in fact have to generate a similar current in the current loop that passes through your toroid, to what you will be trying to measure. Just in case the properties of the toroid become altered at high flux densities. While you could do the tests at a proportionally lower current, it might fail you with a higher current with compression of the output level as the current rises.

The shape of the waveform may have an effect too, if the low frequency component of it results in the current climbing too high in the loop and pushing the toroid into a non linear region.

Coincidentally, I'm just doing up an article on a current sensor circuit using a toroid in a 50kHz vintage computer psu. This information might help you to see how they monitored the high frequency switching output currents. As soon as its done I'll post it.

Hugo.

[trobbins]

Perhaps you could prepare a test current waveform that did reach circa 100A using a charged capacitor, an inductor, a FET switch, and a gate drive circuit that turns off the fet when the target current is reached. You would then have a sawtooth waveform with plenty of high frequency content.

[Al (astral highway)]

Error: signal source is square wave. I posted this while walking along so a bit distracted!

[Al (astral highway)]

Hi Hugo,

It’s a neat coincidence that you’ve been researching on similar lines; I’m looking forward to reading your paper.

Hmm... deep down I did suspect that I’d have to replicate the pulse behaviour of the final circuit. These big toroids are quite expensive at £13 each, so I hope they’re good candidates !

I will put together an IGBT half bridge with suitably beefy devices. I still wonder though that it is a complicated matter to calibrate the current in the circuit. I can assume a certain efficiency (say 95%) for a well designed IGBT half bridge, and measure the DC current, but that gives me average current only. I know I can theoretically use a current sense resistor but in practice , it would be a huge component and wire-wound . So that’s a no go!

The final circuit will be operated at a low duty cycle, in short bursts, at a repetition rate between 60 and 400 Hz. (Repetition rate, not resonant frequency, which will be around 230KHz.) This is an additional complexity for me in terms of making relevant assessments at this stage .

[kevinaston1]

There was an article in Wireless World some time ago about this.

I made a copy of it at the time, so I will have a look and see if I can find it.


Kevin

[kalee20]

With 3.3 ohm burden resistor, you shouldn't need to worry about resonance effects. However, you'll want to minimise the series inductance between the CT and the burden, else you'll have a significant insertion loss.

Any average DC present will act to drive the core towards saturation; this needs to be considered.

Can you sketch out the circuit! From your description, I can't quite visualise it, and the devil's in the detail! In principle, the arrangement is capable of being nearly 'blameless.'

[kevinaston1]

Here is a copy of the Wireless Wold article; it might be of some help to you.

Kevin

[Al (astral highway)]

Originally Posted by kalee20

With 3.3 ohm burden resistor, you shouldn't need to worry about resonance effects. However, you'll want to minimise the series inductance between the CT and the burden, else you'll have a significant insertion loss.

Hello, sure! And actually the burden resistor is just 1R. It's right on the toroid winding. The legs showing protruding are not connected to anything - I haven't cut them as I'll use the resistor again in the next stage - but the scope probe clips on right at the body.

The 3R3 resistor in the diagram is the load applied to the small class B amplifier to require it to draw a significant enough current to pass through the overwind. It could manage 1R5 for the duration of a short test, but I'd rather use an IGBT half-bridge in my next set up.

Originally Posted by kalee20

Can you sketch out the circuit! From your description, I can't quite visualise it, and the devil's in the detail! In principle, the arrangement is capable of being nearly 'blameless.'

Here is a rough visualisation and sketch. The core is huge: 45mm outside diameter, 25mm inside diameter, 2cm deep. It is described a life-size in my sketch but the photo is 30% compressed, so it's bigger than it appears. Hence the path length and X-section area are also gigantic. I think there is no chance of saturating the core, but I will, for peace of mind, do the maths to see how comfortably it could eventually handle 100-300A pulses.

Despite having the 1R burden right on the leads, with <nH of inductance, and Litz wire as the secondary winding, the waveform is very ringy. I do however have quite a long transmission line, which is 20cm coax as well as a short wire, from the amplifier and load resistor to the current transformer.

It is hard to get an average voltage out of the decaying exponential waveform. I was thinking of building a peak level detector - with a difference: that it would sum just the right number of a few of the (larger) cycles of the waveform, exactly such that I get an average figure, but it seems a fiddly and unwelcome diversion as I have other test equipment to put together. And also I had a 'Duh' moment - the 'scope gives me RMS voltage as well as peak, anyway!

There's a blocking capacitor in the output of the amplifier to eliminate any DC in the core. Same will apply for the next, beefier test set up.

[Al (astral highway)]

Originally Posted by kevinaston1

Here is a copy of the Wireless Wold article; it might be of some help to you.

Hello Kevin,

Thank you for digging that out. I think it's a tiny weeny bit cheeky of the authors to call it their design, as it's really just a simple current transformer, but I suppose they are including things like the build details and the connection to the 'scope with the correct impedance transmission line.

It is, however, an ingenious way of avoiding the peril they describe- lifting the leg of the Source/ (MOSFET) or Emitter (IGBT) to attach a wire for measurement purposes and thus blowing the power device by introducing stray inductance which produces fast, destructive transients. So credit indeed to them for imagining how to make a tiny CT that will fit over a leg.

Anyway, I did enjoy reading the article and the exact application is one that I may also copy for another purpose one day, so many thanks!

[Argus25]

Originally Posted by astral highway

I think it's a tiny weeny bit cheeky of the authors to call it their design

One thing I'm not so sure of, with the example bridge circuit in the WW article, they could just have put a current transformer in series with the power supply, which would measure the drain currents for both the conducting mosfets on each half cycle, or better still simply passed the load current through the current transformer and all the important current information is there. I'm not 100% sure why they were interested in the drain currents of individual mosfets in the bridge, when two are always operating in series to pass current to the load, unless to find imperfections in the timing of the drive voltages perhaps or to see if the dead band control (time when all mosfets are off) was correct.

[Al (astral highway)]

Originally Posted by Argus25

One thing I'm not so sure of, with the example bridge circuit in the WW article, they could just have put a current transformer in series with the power supply, which would measure the drain currents for both the conducting mosfets on each half cycle, or better still simply passed the load current through the current transformer and all the important current information is there.

Hey Hugo, yes, well said!

[kalee20]

The current transformer and burden look fine!

Core 45mm OD, 25mm ID, 20mm H gives a cross-section area 200 sq. mm which is pretty big.

What would be helpful is some photos of scope traces - both of what you get across your burden resistor, and what you get across your 3.3 ohm load. These of course should look the same shape, just scaled differently. I'd be worried about DC bias saturating the core, but you say you have a capacitor in series, which will eliminate DC, and that's absolutely fine!

[Al (astral highway)]

Hello; I’ll post ‘scope traces on Monday.

The emphasis willl then be on seeing the behaviour of the toroid with bigger currents.

[Al (astral highway)]

A side-question popped into my head early this morning.

I was thinking that if my 'scope can be set to be triggered in a 'one-shot' mode by instantaneous voltage (lower case v) things are very much easier for me when I'm designing a simple but fool-proof circuit to pass a very high current through a CT. I only need to see what happens for the first cycle of a period of less than 2uS.

I could do with dv/dt, I'm thinking (little v and little t, as instantaneous, not ave.) So I'm thinking the scope does exactly that for a living, only it's gone in an instant, unless it's capable of trigger and hold.

I don't have an instruction manual for my 'scope but I'll have a look at the 'measure' functions and see if something suitable is available. It's a GW INSTEK GDS-1022.

[Al (astral highway)]

Originally Posted by kalee20

What would be helpful is some photos of scope traces - both of what you get across your burden resistor, and what you get across your 3.3 ohm load.

Hey, it took me a while to work out how to test my CT at a higher current. I've dumped the 3R3 resistor, as you'll read below.

I did two trials, one at 250KHz and one in a more complex pulse mode, details to follow!

I binned the original burden resistor as it turned out to be wirewound and hence useless in this role. I replaced it with parallelled 2.2R, 2W 5% metal oxide resistors for around 1.1R burden.


1) The first trial had RMS current of 200mA showing on my bench power supply, but I was driving a square wave signal through a class B follower into a halogen 'DJ' bulb, and the pulses are clearly much higher than the RMS current.

In single-trigger mode, the scope trace showed 200mV. The secondary haas 60T,but for this test, but I had passed two turns over the primary, so the turns ratio is 1:30 instead of 1:60
The 'scope trace is recording a pulse event of 5.5A

2) The second test is with the CT with only 1 turn through the primary. The rig was a 3200uF, 400V- rated pulse capacitor, charged to only 28V (EPCOS, B43415-S9328-A3 type) and then switched into a giant inductor of around 2.6uH (6 turns, spaced. Diameter 4cm, length, 5.5 cm) I calculate the ESR of the circuit (DC, not including complex impedances in the resonant circuit) to be around 0.053R. The cap obviously has to be pulse rated and sees a large potential difference for pulses of <250uS.

In one-shot, 'single' trigger mode, the scope reads a maximum voltage of minus 3.36V.

This corresponds to a peak of minus 200A.

Analysis:

1) The first waveform is irregular, not sinusoidal. This irregularity is explained by the fact that the switch isn't ideal. It's me, manually connecting a wire to the inductor from the charged capacitor (safe, as very low voltage.) The current is large enough to partially fuse the copper in a 1.5mm^2 conductor at the point of contact.

(I wanted a quick and dirty test rather than getting involved with the complexities of power semiconductors at this stage. Calibrating the main feedback mechanism for any future tests is the first priority)

2) There is then a delay of slightly more than one period, 250uS.

3) There is then the peak pulse of around 250uS, sign is minus 200A, showing a magnetic field collapsing in an inductor. The fall time is incredibly pronounced. The -ve going part of the waveform is like a sawtooth but the rising edge is not.

4)then a pulse of around 120uS, same sign, obviously. The capacitor is completely discharged in the initial 'ragged' switch of 220uS, plus the two, resonant events totalling 370uS.

Conclusion: my new current transformer is fine at very high peak currents: thumbsup: I have also calibrated it with a moderate high frequency current (5A approx) and a huge lower frequency current (200A).

This means I can continue with the next bit of test equipment, which will use a semiconductor switch instead of my ragged version.