DIY RF EMC Current Probe Set Design

[dazz1]

Hi
OK now that I am happy with the new test jig, I am back to testing coils.
The latest coil has an inner split floating shield, which is to say it looks like a shield, but isn't.  My theory is that when the inner shield is in very close proximity to the windings, eddy currents form in the shield from the current flowing in each wire-wrap winding.  I think these eddy currents dampen the higher frequency resonance I, Anderson and Pearson see in the coils.

Another feature I have consistently seen is spike around 23MHz.     These looked about right for reflection along the cable connecting the spec-an tracking generator output to the test jig input.  I have been using a 2m BNC cable.
The only difference between the attached plots is the length of the cable.  I used 15cm, 1m, 2m cables and the spike moved frequency with the cable length.

Adding a 10dB attenuator did not affect the spike.
I concluded that the spike is due to impedance mismatch by the coil.  The coil in the fixture is causing reflection of the tracking generator signal along the coax.  Experimentation indicates the spike is a result of e-field leakage through the outer shield through gaps in the foil.  Clamping the 3 parts of the outer enclosure/shield together improved rejection of the spike.

As a side note, the 9 turns on the coil provide an output exactly -20dB from the current passing through the coil (the tracking generator output was -10dBm).  This is a convenient conversion factor to work with.

[dazz1]

Hi
Even with flaws in the foil shielding, the frequency response is good out to 200MHz, but the response changes if the foil is moved.

At the bottom end, the 3dB frequency response is 112kHz.    It is no where near as good as the Pearson coils which go right down to a few Hertz.  112kHz is more than sufficient for my sort of work.   I have the option of making a 2nd coil with a lot more turns to give a low frequency response.

I have now reached the point where foil on 3D printed plastic isn't good enough to get reliable test results.
I need to machine a version from solid metal.    This might take me a few months because I will need a 2mm or 2.5mm thread cutting tap for the screws to hold everything together. 

I think a key parameter is the separation between the coil windings and the aluminium inner shield. If eddy current losses do provide the required damping, then the best way to achieve that is with a multi-strand, non-crossing, flat wound coil.   This will keep the current close to the aluminium.   That is exactly what I have done with the wire-wrap. 

Also the metal to make the coil enclosures is expensive here.  I won't be making a lot of machined prototypes.   I can hold the inner shield with a 3D printed insulating support.  I may be able to do a little experimentation with the size of the inner shield.  I will be aiming to make only one outer shield/enclosure.  I will use a BNC connector if I can, but I think the available bar stock size will dictate a SMA connector.

[dazz1]

Hi
After a search through the workshop, I found some suitable Al bar stock.    The stock is big enough to include a BNC connector, rather than a smaller SMA.
I have drawn up a cad design attached.

The important unknown is the gap between the coil windings and the inner shield.   There is no way of calculating this.   
If the gap is too narrow, the coil will be over-damped.  If the gap is too wide, the coil will resonate.
I have left enough thickness in the inner shield to increase the internal diameter to widen the gap.
If I go too far, I can reduce the thickness of the entire assembly to reduce the gap on the top and bottom of the coil. 

What is not shown is that I will be using perforated overhead projector sheet as an insulator between the various metal parts. 

The internal diameter of the through hole is only 22mm, which is not large but sufficient for my needs.   The offset hole simplifies manufacture. Very little milling will be required. 

The plan is to make this over the next couple of weeks.

[dazz1]

Hi
I have started machining the inner shield, which looks like a shield but isn't.  It is fully floating.  It does act to dampen resonance.  I think it works by the coil flux leakage inducing eddy currents in the aluminium.  The effect increases with frequency, which is good because resonance occurs at higher frequency.

I should note that the design is terrible.  It requires trepanning a deep groove into the face of the bar.  The coil is then pushed in.  This will make removal of the coil very difficult.  I did it this way because it uses the minimum length of bar stock.  That stuff is expensive.

[mag_therm]

Nice trepanning job, I have done that on  medium format camera lens adaptors.

I had a failure of the little HF 30 T c/t here. I might be lucky it was  not on my new 'scope!. The old Eico 410 10 Mhz  survived.
The attenuator is ~ 4:1
Previously:
R11=R22 = 50 Ohm 1 watt precision
R21=R12 = 120 ohm 1/4 Watt.

I had the c/t on primary of inverter transformer
When the inverter started into low load, the transformer momentarily saturated ( separate problem)

Resistor R11 blew to pieces and started arcing due to the higher resistance (290 Ohm) seen by c/t

So I changed to
R11 = 56 Ohm 3 Watt paralleled with 470 Ohm to get near 50 Ohm
R22 still 1 Watt precision but with  but added paralleled back to back zeners 4.7 V 1 Watt .
I will make a board for that with a semicircle to glue the ferrite core and bnc onto.

I am using these as pre-wound c/t, need to get some more @ $0.50 ea
Don't know what the core is and they work as transformers up to 100 MHz, I used one as UnBal for the FM broadcast receiver
https://www.surplusshed.com/pages/item/M2199.html

Edit: ordering some of these for the burden/attenuator:
CGS 355051RFT
SMD Chip Resistor, 51 ohm, ± 1%, 5 W, 4320 [11050 Metric], Thick Film, High Power

Max 300 V  rated working 111 Volt
Maybe that size not necessary for RF use

[dazz1]

Hi
A little more progress.  The core and coil are fitted in the internal shield.  Just a reminder that the coil is 9 turns.  Each turn is 6 strands of wire wrap.
I drilled a hole for the coil wires out the side.
I drilled 4 holes in the back of the inner shield so I can push out the coil and core if needed.
The metal on the internal bore is 0.5mm thick to maximise the hole diameter.  It also has a tapered knife edge to reduce the capacitance between the cover.
The thickness of the external metal is thicker than needed so I can adjust the gap between the windings and the aluminium if needed.

[dazz1]

Hi
I did some testing with the inner shield.  It look promising enough to go ahead and make the (actual) outer shield.
The peaks under 100MHz are due to reflections down the cable.  Changing the cable length changes the frequencies of the peaks.  This is shown in the differences between the two overlaid traces.  The two traces are different colours.

The lower 3dB point is about 125kHz.  Not even close to the Pearson specs, but adequate for my purposes.

The 3dB bandwidth is at least 125MHz, usable with corrections out to 200MHz.   This is reasonably close to the results seen with the plastic and foil prototypes.

Tests with the prototypes indicate that the frequency response should flatten out once the outer shield is fitted.  I hope that is the case with the solid aluminium version.

I am not going to try to tune the response.  I will just go ahead and make the outer shield and take it as it comes.  It already performs far better than I expected it would when I started.

[dazz1]

Hi
Some more progress.

The main body of the current probe features the really bad design requiring trepanning a deep groove for the coil to sit in.

I did the poor mans version of anodizing by boiling the inner shield in water.   I discoloured to something that could easily be mistaken for titanium.  It will never be seen once probe is assembled.

I made a sacrificial fixture from a round piece of 6mm steel.  The probe is clamped onto the fixture so I can hold the work machine the various surfaces in one pass.

The fixture has the holes for the screws marked on it.

Everything is clamped together to drill the holes for the screws.  This will ensure all of the holes are perfectly aligned.

If this probe doesn't work, it will be a very expensive and plain looking paper weight.

[dazz1]

Hi
Photos of the current probe ready to drill the holes for the screws.
I will use 6x 6BA screws, because more screws give the impression of better quality, and because I have them in the drawer. 

[dazz1]

Hi
Some more progress with milling out the pocket where the BNC connector fits.
Also I milled the flat where the BNC fits.  This was done with the cover screwed on to ensure a perfect join between the cover and the main body.    This is why the cover has six screws.  It had to be strong enough to survive being milled in place.

[dazz1]

Hi
Today I milled out the hole for the BNC connector.  The threaded part is round with two flats.  I wanted to minimise and twisting of the BNC connector when in use.  This required a close fit and a hole that was far from round.
I used a CAD programme to figure out the milling way points to approximate the hole shape/size I needed.     I created a table of x/y coordinates I needed to reach.  CAD told  me I would need to file off some of the high points.   The milling machine didn't listen.   The hole ended up being the exact right size and shape.  No fettling was required to allow the connector to slide into the hole.

[tautech]

Hi
Today I milled out the hole for the BNC connector.  The threaded part is round with two flats.  I wanted to minimise and twisting of the BNC connector when in use.  This required a close fit and a hole that was far from round.
I used a CAD programme to figure out the milling way points to approximate the hole shape/size I needed.     I created a table of x/y coordinates I needed to reach.  CAD told  me I would need to file off some of the high points.   The milling machine didn't listen.   The hole ended up being the exact right size and shape.  No fettling was required to allow the connector to slide into the hole.

Nice work Dazz. 

[dazz1]

Hi
I completed the final machining operation by turning down the external dimensions to size.
With the enclosure screwed onto the fixture, I was able to machine the face and sides in one pass for best finish.

Now I just need to assemble and test the current probe.  Assembly includes cutting insulating sheet material (clear photocopy film) to separate the various parts where required.
I think testing will be the most stressful task because of the risk that the whole thing is a failure.

[dazz1]

Hi
OK so I have finished assembly and testing.

Attached are images of the final current sensor compared to the original CAD cross section.
The machined current sensor all worked out as planned. I didn't have any disasters.  Everything fits.  It has a nice shine.

[dazz1]

Hi
I measured the performance.  It was better and worse than the 3D printed versions with kitchen foil.

It was better in that the resonance peak at about 23MHz is no longer there.  This peak was related to the connected cable length and was not suppressed by an in-line 10dB attenuator.  At the time my hypothesis was resonance was due to e-field pickup through gaps in the foil.  The final version was fully e-shielded has no resonance peak. 

The frequency response is worse in that it is smooth but not flat.  The prototype showed a flat response out to about 80MHz.    The final version has a significant slope.  This can only be due to the aluminium enclosure.  The hypothesis is that the leakage flux from the coil windings is generating eddy currents in the aluminium, and these eddy currents cause losses.  In effect, this forms a frequency dependent resistor.  The higher the frequency, the lower the resistance.    The positive benefit is that resonance of the coil is fully suppressed.  It is clear that the damping is too high causing a relatively straight slope.    The cure would be increase the clearance between the aluminium enclosure and the coil windings to reduce the damping effect.  That is not a trivial task and I am not going to do that.

The mitigation is that my spectrum analyser produces numerical CSV files that I can apply to measurements to correct for current probe errors.    The good thing is that the frequency response is devoid of resonance which would be injected into the circuit under test, if it occurred.   Too much damping is better than too little.

So looking back at what has been achieved.
I developed a current source that is in effect an open coax, which is easy to make and demonstrates good wide band frequency response.
Proved that a 3D printer and kitchen foil can be used to produce reasonable prototypes for this application.
Eddy currents are an effective method of damping coil resonance
It is possible and practical to DIY build a current sensor coil up to at least 200MHz
Improved low frequency response would require a nano-steel core.  Not easy or cheap to source.
Optimising the design of a current sensor with eddy damping would require experimenting with the clearances between the enclosures  and the coil windings.  That would require a significant R&D effort.

I started this project hoping to get up to 30MHz for EMC testing.   I have ended up with a working current sensor with a usable range >>30MHz but with sub-optimal performance.  There is a lot of room for improvement, which the prototypes showed is achievable.

[dazz1]

Hi
Today I did some investigation and experimentation trying to figure out why the 3dB bandwidth is down to 60MHz instead of the ~160MHz expected.  I don't even need 60MHz but I'd still like to know why.

I checked the current fixture.  Even though the current probe is a tight fit within the fixture, it still works very well.  I figured out that if I connected a 10dB attenuator where the 50ohm load is normally connected, I could normalise the injected current with the spectrum analyzer.  I could then remove the small error of the fixture (<0.5dB).

Analysis of previous measurements isolates the reduced performance to the outer shield/enclosure.  There is something about the outer enclosure that is causing a significant loss, rising with frequency.  It can't be the solid metal sections because at 100HHz, the skin depth is only ~7um.  I was able to machine the outer enclosure to increase the gap to the inner shield.  That had no effect. 

I checked that the inner and outer shields were not shorted together.  No short and no change.

The design offset the radial flux breaking grooves between the inner and outer shields. This offset improves e-shielding and eliminates any e-field break through.
I tried aligning the two grooves but this only allowed a small and visible break through.

I don't know what is causing the significant increase in losses as the frequency rises.  Based on development with the 3D printed prototypes, I was expecting a flat response out to about >150MHz. 

The first image shows the frequency response of the fixture with the current sensor fitted in place.  The fixture is good out to 600MHz, and usable well beyond that.

The other plots show the frequency response of the current probe.   The 3dB point is at 60MHz, well above my initial requirements for EMC work, but still not as good as the 3D printed prototypes indicate should be achievable. 
The current probe is entirely usable out to 100MHz to make relative current measurements.  Absolute measurements will require error correction.

The photograph shows the setup to measure the fixture error with the current probe fitted in place.   The current probe is terminated with a 50ohm load to simulate connection to the spectrum analyser.  This load will appear as a parallel load of about 450ohm to the current going through the fixture.  The 10dB attenuator terminates the fixture current to 50ohm.   The output of the attenuator measures the actual current passing through the current probe, which includes the effects of the current probe in the fixture.  This is a good way of measuring the performance of the fixture and also the influence of the current probe on measurements.  Having a big chunk of metal in the middle of the fixture has little effect on performance.    It is also simple, easy and cheap to make.

So in conclusion, the current sensor exceeded the original target of 30MHz bandwidth for EMC work but the 60MHz achieved on the final version is substantially less than the  +150MHz bandwidth indicated by the 3D printed and foil wrapped prototypes.  There is too much attenuation as frequency increases.  Resonance is heavily suppressed, which is good, but sensitivity also suffers as frequency rises. 

The double skinned design completely eliminates e-field break through.   There is no pickup of stray e-fields.  This is good because the current sensor is only measuring the current through the probe, and not the voltage.

So I now have a nice shiny paper weight that can measure currents out to >60MHz.  More R&D would be required to figure out why the outer enclosure is reducing the bandwidth, but I am not going to go down that path. 

[garrettm]

I wanted to say that I'm impressed with your workmanship dazz1!

That said, I think the shield should be split near the center of the coil (rather than at the top edge) and a split should be added along the top and bottom plates (leaving only the outer edge as a continuous strip of metal). This is how the clamp type wideband current transformer I have is constructed.

[dazz1]

Hi
I experimented to see if the position of the slot made any difference.  I saw none.  I did find that with just one slot, there was e-field leakage that appeared as a small resonance at around 20MHz, depending on the length of the cable between the current sensor and the spectrum analyser.  Inserting an attenuator did not affect the resonance.

 I think it is more likely that placing the slot in the centre allows for two identical parts to be made for the e-shield.  They would have been spin-formed.  I did consider using metal spinning to make two halves of a shield but I would need to make a form.    The other problem is that I would have needed to come up with some really ugly arrangement to make the connection to the BNC, just like yours.      I decided if I was going to make a paper weight, it had to look good.

If you have a look through the posts, I tested the finished un-earthed inner enclosure before I even started making the outer shield.  It had heaps of bandwidth but also lots of e-field.  I am just wondering if the bandwidth I thought I saw was the bandwidth of the e-field leaking down the coax shield.    Maybe the "perfect" e-shield of the metal outer enclosure simply revealed the underlying H-field bandwidth of the coil.      Maybe the e-field and h-field added together to give the flat frequency response I saw. 

The other issue I have is that the coil and core over damp resonance.  That could be core losses.  I don't know the history, make or material of the core.  It is a junk-box special.  That can't be eliminated as a root cause.   Over-damping is much better than under-damping.

The one thing I have not yet tried is testing with the inner enclosure not fitted.  I might get around to trying that but I already know that will increase e-field leakage and might give false hope.

Given that my original intention was to DIY a 150kHz to 30MHz EMC current sensor coil, making a 60MHz coil doubles that target.  I have met the requirements.  It is a great looking paperweight that can also measure RF conducted current.

The current fixture is a success and I am considering making slightly larger, improved version, to normalise the current sensor with the spectrum analyser before each use.   This will take out all of the errors.

[dazz1]

Hi
Based on my test results, I suspect there may be sensor coils out there that fake their bandwidth because the coils are sensitive to both e-field (voltage) and h-field (current).

As standard, sensor coils are calibrated with 50ohm source and load impedance.    There is a simple test that could be done.  Change the impedance seen by the coil as shown in the attached diagram.  If the coil is only sensitive to current, then the current should be reduced in proportion to the increased impedance.  If the coil is sensitive to e-field, then the output will increase based on the higher voltage.

I think it is entirely possible that simpler, lower quality sensor coils might be able to claim a higher bandwidth than a fully e-shielded sensor coil.