DIY RF EMC Current Probe Set Design

[dazz1]

Hi
I have read posts on EEVblog Forum, and papers elsewhere that describe how to DIY build a RF current probe (like these professional ones https://www.fischercc.com/type/current-monitor-probes/ ) for EMC work.  Most of these posts and articles are quite old and describe one size.

I would like to build a set of say 3 different sizes specifically form EMC work.   I'd like to use the latest materials and techniques to make professional grade probes at DIY prices.
The reason I want to build a set is because  I don't think one size will fit all situations. 
Small sizes can be expected to have higher frequency range but lower current limit.   A small size could be easier to use.
Conversely, larger probes will have a lower frequency range but higher current limit.   The increased physical hole will be useful.

Is there anyone interested in participating in the design phase? 

[jonpaul]

See Tektronix Circuits Concept book Oscilloscope Probe Circuits for current probe theory and design.

You can buy a fine Tektronix P6022, perhaps $50-150.

larger P6021 is similar

We used Pearson 411 toroids for wideband

As you can wire in current sense jumpers, probe sizes are not a problem.

Jon

[dazz1]

Hi
I looked up the P6022 an new price is about $USD3800 https://nz.element14.com/tektronix/p6022/probe-current-6a/dp/7986882.  e-bay prices are around $USD1000.
The Pearson 411 coil is about $USD900 including import tax and shipping. https://www.bcgroupstore.com/Biomedical-Pearson_Pearson_411.aspx 
Waaaay too many $$$ for me.

[jonpaul]

TEK probes: Epay ...used prices but be patient. Probes must have termination box.

Best laces at ham Radio flea markets.

DIY is hard for clamp on, easier for toroid.

Jon

[tautech]

daz, why would you want a current probe for EMC testing as opposed to a near field probe ?

Would you not be using your SSA for this work ?

[2N3055]

daz, why would you want a current probe for EMC testing as opposed to a near field probe ?

Would you not be using your SSA for this work ?

Current probes are used for conducted interference.. They are basically very wideband current transformers..

[artag]

They're not that wide. P6022 is good to 120MHz, 6021 to 60MHz.
Conducted interference frequency limits are lower than radiated, but I'd regard that as a bit low for EMC work.

[jonpaul]

Bonjour,

the conduced EMI range in EU and UK, US is 100k..150 k >>>30 MHz

See EN 55032 for Class A and Class B conducted emission limits

https://e2e.ti.com/blogs_/b/powerhouse/posts/a-review-of-emi-standards-part-1-conducted-emissions
https://metlabs.com/emc/en55032-replacing-en55022-multimedia-equiptment-march-5-2017/

in 1970s...1990s I  used  P6021, P6022 and  TEK probe amplifier 134, also  Pearson 411 toroidal CT in conducted EMI lab work.

An appropriate LISN and calibrated spectrum analyzer are essentials for conducted work.

 The  conducted EMI is reduced by design: soft switching, transformer shields, CM, DM filters.
After testing, one could then  mod  PCB layout, add cable and PCB shields and ferrite beads.

Above 30 MHz EMI  is radiated and NOT not conducted. Radiated is NOT read with current probes! Use  a sniffer for E or H field.

Bon Chance

Jon

[tautech]

Bonjour,

the conduced EMI range in EU and UK, US is 100k..150 k >>>30 MHz

See EN 55032 for Class A and Class B conducted emission limits

https://e2e.ti.com/blogs_/b/powerhouse/posts/a-review-of-emi-standards-part-1-conducted-emissions
https://metlabs.com/emc/en55032-replacing-en55022-multimedia-equiptment-march-5-2017/

in 1970s...1990s I  used  P6021, P6022 and  TEK probe amplifier 134, also  Pearson 411 toroidal CT in conducted EMI lab work.

An appropriate LISN and calibrated spectrum analyzer are essentials for conducted work.

 The  conducted EMI is reduced by design: soft switching, transformer shields, CM, DM filters.
After testing, one could then  mod  PCB layout, add cable and PCB shields and ferrite beads.

Above 30 MHz EMI  is radiated and NOT not conducted. Radiated is NOT read with current probes! Use  a sniffer for E or H field.

Bon Chance

Jon

Yes for conducted EMC P6022 could be an excellent choice and I'm a fan of these old designs....6022 and 6021 both with terminations in my selection of equipment. I also have a Type 134 and 110&230VAC supplies for it.
Like you said, with patience you can find these on eBay for 2-300 if you wait.

The OP has IIRC a SSA3021X that does have a EMI mode but not with the preloaded CISPR ranges like the later SSA3000X Plus models however you can configure the OP's SA for the appropriate levels for radiated EMI.

[dazz1]

Hi
I want to build up my tool set for EMC work.    I have probes and I have a couple of log period antennae by WA5VJB that cover the rang 400MHz to 6.5GHz. 
My Siglent SSA has all available options enabled, including EMC and 3.2GHz freq range.
I built a directional coupler.
I have already purchased Fair-rite type 61 toroids (plus a larger unknownium toroid) to experiment with.
I have recently repaired an antique Wavetek 2520 RF sig-gen that gives me a quantified signal source.
I have a bi-conic antenna on the drawing board.


What I don't yet have is any Litz wire, but I would save that for the final versions.  For prototypes, I plan to use multiple strands of wirewrap to approximate Litz wire.

With what I have, I am aiming to produce at least two current sensors.  One with a bandwidth up to 1MHz-5MHz, and the other up to 30MHz (if that is even achievable). 
The specs for commercial units indicate that you can get sensors with a good lower frequency response, and sensors with a good higher frequency response, but not in the same coil.

One issue that I haven't seen in any documents I have found, is the effect of the coil on the circuit. You can't measure something without having some sort of effect on the something.   I would anticipate running a cable through a toroid will tend to attenuate any signal on that cable.   This would apply even if a LISM is used.   The risk is getting false confidence that the conducted emissions are less than they really are.

If the toroids work well, I might consider making clamp-on versions to give more flexibility. 

[tautech]

One issue that I haven't seen in any documents I have found, is the effect of the coil on the circuit.

Better tools do list Insertion Impedance.
See P4 of the attached Tek P6022 current probe manual.

FYI the P6021 manual is also attached.

[jonpaul]

Bonjour Dazz1: fine work, a few points....

1. Wideband transformers and CT toroids  are very difficult to design and build,   Pearson 411 current transformers  are optimized for  :-BROKEvery low freq to 50 MHz and cost $500-800.

2/ CT TEK probes LF end can be extended with the 134 CT amp.

3/ DC hall effect probes with a special plugin or amp can achieve DC-50 MHz.

4/ See other posts on this topic for more info on EMC probes.

5/ Litz wire is not needed for the CT.

Jon

[dazz1]

Bonjour Dazz1: fine work, a few points....

1. Wideband transformers and CT toroids  are very difficult to design and build,   Pearson 411 current transformers  are optimized for  :-BROKEvery low freq to 50 MHz and cost $500-800.

I have no intention of trying to make a Tektronics type probe.   

2/ CT TEK probes LF end can be extended with the 134 CT amp.

I have already built a rf-preamp that is powered from the USB port on my spectrum analyser, or any other 5V supply.

5/ Litz wire is not needed for the CT.

I agree it isn't needed, but I can make some up with my existing stock of silver plated wire wrap. 

[mag_therm]

Hi Dazz,
Some tips for a 1 Turn solid toroidal C/T design.

Select a turns ratio and burden resistor for easy conversion eg 1V/A, 2mV/mA etc.
The burden resistor should be of suitable precision, non inductive, and  has to be rated for low temp rise to help precision.

For a Ferrite core, initially aim to run B_peak at 0.1 T peak at lowest sq wave, and down to about 0.02T at highest frequency.
You may be able to do a better range with testing.

As the voltage is measured into a diff amp across burden resistor, the winding resistance does not affect accuracy.
However it does effect the B_peak and core temp rise, so you need to calc or measure winding resistance over the frequency range.

The only times I used litz wire was in the C/T's measuring up in 5,000 ~10,000 Amp range which were water cooled and winding and core loss had to be minimized.
Your cores will hopefully have very low temp rise. As ferrite cores are usually optimised for around 100C, you need to get from data sheet, the core characteristics at ambient temperatures which are usually not as good as when hot.

Good Luck ! Post some details as I for one, and sure jonpaul too,  will be following your development.

Edit: And should mention the low level IMD and distortion at very low flux density. I never encountered it in power electronics, but did in some hobby 3~30 MHz antenna tuner inductors.
Search: Peterson coefficient, Legg coefficient, and Rayleigh loop. Distortion is a reason not to run down at very low flux density.

[dazz1]

Hi
I made some progress today.   I decided I needed to make a test fixture to be able to get consistent (not the same as accurate) results.  I have not seen a commercially made test fixture, so I designed my own.
I used a U bent piece of aluminium sheet to provide an earth return and to physically hold things in place.  The coax centre wires are connected with a piece of bronze brazing wire, soldered to the BNC pins, that is cut close to one end.  To fit a coil, I remove the  BNC connector with the longer brazing wire.  Fit the core in-place and refit the BNC connector.  The terminal block then connects the two ends of the brazing wire. 

I have one side of the fixture fitted with a 50ohm load.  The other side is connected to the Tracking Generator output of my spectrum analyser. 
The coil ends are connected to coax that goes to the RF input of my spec-an.   The centre of the core is stuffed with a piece of foam to keep the brazing wire centred through the core.   The aim being to get consistent repeatable results. 
I have no idea if this is the "right" test setup for RF current sensors, but I have used what I have available. 

I made a coil with the unknown core and randomly chose to do 10turns with some crappy scrap hookup wire.    I had really low expectations but I was wrong.

If you take a look at the spectrum analyser screen shots, this coil has a flat response out to 21MHz.  There is a resonance spike but this is sensitive to the positioning of the coil ends and the coax connections.  With not much tweaking, this spike disappears. 
This coil has a useful response out to about 150MHz, far higher than I expected. 

On the low frequency end, the frequency response is also much better than expected. 

To test whether the results are actually real, I removed the 50ohm load.  As expected, the frequency response went nuts and was full of spikes and resonance as the rf from the tracking generator reflected back.   It means that the coil is responding to what is passing through the centre.  So I am reasonably comfortable that what I am seeing is real. 

I will have to use a DC current source to figure out the saturation point.  The equipment I plan to test draws less than 1A @ 220VAC so I don't expect any issues.  I need to add some electric field shielding and do some more testing. 

These initial results are good enough to justify further time and effort to do more R&D. 

[jonpaul]

Daaz1,  BRAVO for your fine efforts and progress.

We used wideband CTs for HID and arc lamp ignitors and SMPS designs,  decades ago 1980s ..1990s
We had contact to Mr. Pearson, a pioneer in the field.
A few moments for a trip down memory lane...

1/ Cores are high perm, besides 10k..15k ferrite, I believe thin tapewound cores and amorphous tapewound cores were used.

2/ Application with high DC bias need a gapped core or powdered iron.

3/ Commonly the problem is sidestepped with a DC bucking technique on a second primary.

Thus the core sees only the AC component, the DC component  is cancelled by the buck winding.

A TEK high current clamp on extender CT-4 comes with an accessory  DC buck coil.

https://w140.com/tekwiki/wiki/CT-4

4/ Winding is sectionalized and a distributed termination is used. Each section has a term resistor.

5/ Coil term is optimized for best BW or HF transient response, independent of required system Zo eg 50 Ohm.

6/ The coil term is then adapted to the BNC or connector Zo with a matching network.

7/ Calibration constant is adjusted in the term and matching network eg 100 mV = 1 A

8/ Number of turns increase with required LF corner.

9/ You can test at higher excitation with a multiturn primary eg 10 T multiplies the primary A*T by 10.

10/ The finished toroid is shielded against EMI and capacitive noise by a metal shield which has breaks to avoid shorted turns.

11/  occasionally can find a used Pearson 410, 411  at Ham fleas, surplus, or epay. Patience and persistence.
We have several,   paid $25.....$75 over the decades.

12/ I think the design details are in old fine papers, app notes and patents from Pearson and others.

Just the ramblings of an old retired EE.

Jon

[dazz1]

Hi
I think I still have a lot of work to do.

e-field shielding and resonance
I suspect putting this coil into a metal shield will add capacitance between the windings and ground.  The capacitance plus the stray inductance from the coil windings is a recipe for a tuned cavity resonator.  I expect that will really mess up the frequency response.  The problem with typical shielding is that it doesn't absorb the stray energy. It is like yelling in an empty concrete room.  The sound reverberates and echos.  If the room is lined with absorbent material, the sound dissipates. 

You can buy specialist EMC RF absorbent material that sticks onto the inside of shielding to absorb stray RF.  Naturally this is difficult and $$$$ to source.  I glue a layer of very small ferrite beads onto the interior surface of shielding.  Grade, size, shape etc is not important.  I can get 100x for $1 from China if I am patient.  I focus on corners and edges where currents are a max.    This method has worked for me.

Matching
I haven't even looked at matching the coil to the 50ohm source.  With a 10:1 turns ratio, the 50ohm load on the through wire is going to be something less than 10ohms through the coil.   I anticipate the coil is going to look like a mostly highly reactive load (it's an inductor), so matching will require reactance.  Achieving wideband matching would be a challenge to say the least.   This would be further complicated by the effects of the measured impedance reflected through the coil.   At present the SA Tracking Generator and fixture load are 50ohms but if I replaced the 50ohm load with a capacitor, the coil impedance is going to look quite different.    I don't have a network analyser to fully characterize the coil.  I do have a RF bridge to do some limited measurement.  This only gives me amplitude, but not phase information.

I am going to try varying the turns to see what effect that will have on bandwidth, gain etc.  At this stage, I won't be trying to match the coil to the spectrum analyser input.

Litz Wire
The frequency response shows a downward slope, most likely due to loss in the ferrite or the coil.   I suspect the increased resistance due to skin effect in the cheap hook-up wire will be observable by comparison with the Litz wire.   

I am going to try some DIY Litz wire, using wire wrap, to see if that makes any difference.

Pearson
  The range of used electronics test equipment available here is pitiful and expensive.    There is absolutely no danger of finding a cheap Pearson in my country.   I look with green eyes of envy at the range of cheap test equipment on USA e-bay but the shipping costs rule out this option.  Also, I hate dealing with e-bay.   e-bay has no presence in New Zealand because it was wiped out by the local competition www.trademe.co.nz  Going to www.ebay.co.nz redirects to ebay.com and no local listings.

There is a really nice HP Network Analyser for sale here  https://www.trademe.co.nz/a/marketplace/electronics-photography/other-electronics/other/listing/3646247523?archive=1&bof=WJH0nzdg but the asking price is eye watering. 

[jonpaul]

Dazz1:

I Understand now your issues with epay, etc.

Perhaps contact an NZ or AU Amateur Radio club, or ARRL branch for ham fleas....

The Shielding of the Pearson's is simply a metal clamshell split in both axies, for E field electrostatic shield.

Litz is very specific, many very fine strands with dia ~ skin depth, interwoven not just bunched or twisted.

https://en.wikipedia.org/wiki/Litz_wire

You may find some on old RF chokes....just unwind.

As the currents are so small in these CT toroid's, i cant imagine use of Litz makes any difference ....I think the Pearsons used regular solid magnet wore, not Litz.

Bon Chance,

Jon

PS:  Had no idea you are in NZ, the EEVBlog user  flag icons are soo small they are often unrecognizable....

[mag_therm]

That is a good looking test jigger you made.

If using a rounded toroid, I wind enameled wire (thin insulation) directly on the ferrite, spaced evenly around the core.
The current flows biased to the side of the diameter closest to the core. ( In fact rectangular turn section would be more ideal approach to a "current sheet")
If using thick insulation, or Litz,  I have not measured, but an airgap between winding and core implies leakage flux.

The winding is indeed an inductive device, but the way a current transformer works, is that the secondary Amp.Turns is "exactly" the same as the primary.
 That assumes mutual inductance factor k = 1. The secondary current is not dependent on frequency, up to the limits of the low frequency model.
So you need short leadouts to non inductive burden - which I see you will have, with 50 Ohm RF input.

P6021
The circuit diagram for the P6021 passive matching box seems to be published.
(I know because I have copied into my notebook years ago. We used many these probes exclusively for FET, IGBT and SCR gate current measurements. The probes retain accuracy even in the presence of DC, as in scr gate pulses, the scope trace balances to the average DC )

P6021 burden  is a 68 Ohm shunted with two low pass filters plus some other components. The turns are 125 ( written on the side), the net burden is close to 62.5 Ohm. There are some frequency comp adjustments on the probe head.

So for 200 mA measured,  , the winding current is 200/125 = 1.6 mA
The burden voltage going to high impedance scope input is 1.6 * 62.5 = 100 mV
That is how the ratio is obtained ( 2mA / mV ) in the case of the P6021.

The core area (visible) is about 2.5 by 4.5 mm, for nett core area approx 10e-6 m^2
For the current measured  of say, 2000 mA rms , 120 Hz is the lowest rating, worst case for B_peak. Highest rating is 60 MHz
E_burden = 2000 *(1/2)  = 1000 mV
E = 4 * F * N * B_peak  * A
1.00 = 4 * 120 * 125 * B_peak * 10e-6
B_peak = 1.6 T
Most high frequency pulse measurements will be fractions of that value.
For example  2 Amp 10 usec pulses @ 50 kHz, B peak = 4 mT approx

[mag_therm]

Hi Daz,
I made a c/t with 30 turns of 24AWG on a 30mm ferrite toroid.
The 50 Ohm burden was close to toroid, then cable to HP141T SA

Just on wood block, it has about -70dBm of unwanted capacitive coupling and  also inductance off the bnc connectors to the brass tacks.
HP141T only goes to 110 MHz
Measuring from 1MHz to 99MHz with -10dBm  ( 1.414 mA rms) from sig gen  to the 1Turn input.

The calculated output across 50 Ohm burden is  -39dBm
measured:
The output is  -40 dBm from 1MHz to 30 MHz, then a hump up to about -35dBm from 40 to 60MHz then flat at -42dBm again to 99MHz.

I could make a test jig like yours, and I think a good hi Z diff amp is needed close to the resistor to reduce the common mode coupling,
and lead inductance.
https://app.box.com/s/s333n05r6pktw4uc6yb6jjuxom1e69xv

[Wolfram]

A lot of the fancy tricks with complicated compensation networks, distributed termination and nanocrystalline cores are utilized to get as wide bandwidth as possible, for example the discussed Tektronix P6021 has a bandwidth of half a million to one, and the Pearson 2878 has a bandwidth of over two million to one. I have several of either, but I find for EMI measurements something simpler will work just as well. A plain resistively terminated current transformer wound on a ferrite core will easily give a bandwidth of ten thousand to one with some care in the design, plus minus 3 dB that is. It doesn't beat a five hundred dollar Pearson that comes with a calibration report, but for EMI precompliance testing it's often more than good enough to be useful. There are even commercial EMI measurement CTs made with ferrite cores and lumped termination, like the Tekbox TBCP1-200 for example.

To get a predictable sensitivity, you want to terminate the sense winding into a (non-inductive) resistor. The resistance along with the magnetizing inductance of the secondary gives you your low cutoff frequency, forming a single pole where the magnetizing reactance equals the terminating resistance. The resistance together with the number of turns gives you the sensitivity of the current transformer. More turns gives more parasitic capacitance however, compromising the HF cutoff of the coil. High end wideband transformers from the likes of Pearson, Stangenes and Bergoz tend to use very high permeability cores made from strip-wound nanocrystalline material, to increase the magnetizing impedance as much as possible without compromising HF performance by having a large number of turns.

As an example, a Fair-rite 5975002701 core has an Al value of 5500 nH/n^2. If this is wound with 10 turns and terminated into 10 ohms, you will get a sensitivity of 1 V/A and a magnetizing inductance of 550 uH, giving a lower cutoff frequency of 3 kHz. The upper cutoff frequency will have to be determined experimentally, but a few tens of megahertz is not unrealistic. To match this to 50 ohms, you could put 40 ohms in series with the output. If you have a way to generate a test signal and monitor the current transformer output, I would highly encourage doing some experiments.

[dazz1]

Hi
Today I wound the unknown type core ( I have two of them) with 9turns of wire wrap.  Each turn was made of up 8x wire wrap as DIY Litz wire.  The primary aim of the experiment was to measure the benefits (if any) of using Litz wire.
The wire wrap was wound as a single layer with even spacing.  Spacing the wire wrap minimises proximity effects.  Winding the wire wrap close to the core minimises stray leakage.

The results are attached.
With rubbish hook up wire, the frequency response degraded with frequency.  The hypothesis was that this was due to ferrite or copper losses. 
The use of DIY Litz wire shows no degradation of frequency response attributable to losses. 

The measured 3dB bandwidth is from 111kHz to 170MHz.    There are no nasty spikes in the response.  For most of the bandwidth, the response is within 1dB.  Beyond the 3dB upper frequency, the coil is still usable.

It seems reasonable to conclude that: 

• using Litz wire provides measurable improvements in frequency response.
  
• hand winding Litz wire is surprisingly difficult.  Next time I think I will make a simple jig to help.
  
• the performance of this coil as a current sensor is well beyond expectations.
  

I will try more windings to improve the lower frequency response.
I will also start looking at making a metal e-field shield enclosure.  I will be specifically looking at the effects on frequency response (cavity resonance).

If I see problems with cavity resonance, development will effectively be stopped for a few months while I order, then wait for a bag of ferrite beads to be delivered from China.
They are available locally but are sold individually.  Buying 500 or so would cost more than the national debt.

[mag_therm]

Looks good.
Litz wire in the past referred to strands woven so the each strand passes  a period on each side and inner  of the core ( my bad description.)
From your photo , that winding used to be called "bifilar.. trifiar...multi-filar" where the strands are each in parallel flat on the bobbin or core. That is a good way to approximate a current sheet which reduces leakage ( ie keeping flux in core, not in air), and also keep strand currents balaced.

I suppose your next test could be to run at various primary levels from low to high dBm  to check  the "gain" and phase over the range you need.
I don't think I can do that accurately here, I can only go to +10dbm and 'scope  is limited to 50 mV/cm @ 10:1 probe.

[dazz1]

Hi


I don't have access to the exotic materials that might allow me to make a single really wide band current probe that I don't actually need.  Two probes that cover the required bandwidth would be sufficient.     
I am also planning to make a lower frequency version.    I can fit 160 turns of wire wrap on a single layer.    If each winding is made of four strands of wire wrap, then I can make a current probe with 40 turns.

Most commercial EMC standards have measurements made over the frequency range of 150 kHz to 30 MHz.  I have already made a current sensor that measures over that frequency range so I could stop now.     

A lower frequency coil could then be used with my oscilloscope.   Rather than using a burden resistor, I would use a 3dB attenuator to act as a burden resistor and matching to the coax.

Before I test any more coils, I want to test a prototype enclosure.  I have designed an enclosure that is like a donut with the hole offset.   The offset will provide the space for the SMA panel connector.   This shape will be the easiest for me to make because it only requires a lathe.  I have a milling machine but milling out an enclosure would take a lot more time than a lathe.  The plan is to print a 3D plastic prototype , and line it with aluminium foil for testing.  If that works without wrecking the frequency response, then I will machine an enclosure from aluminium bar stock .

[dazz1]

Hi
Today I tested the 3D printed, aluminium foil lined enclosure.  As expected, it definitely affected the frequency response.
There is clear evidence that the enclosure acts as a tuned resonator. 

The Spec An shows two conditions:
1.  The coil is electrically isolated from the aluminium shield.  This is not a "normal" connection.  The effects on the frequency response are measurable, but too not bad.

2.  One side of the coil is grounded to the aluminium shield.  In this condition, the enclosure clearly acts as a cavity resonator pushing the frequency response down by 40dB.  This definitely affects the upper 3dB bandwidth of the coil.

In both cases, the frequency response gains bumps and dips due to the presence of the enclosure.

One of the plots below has two traces, one with the coil grounded to the enclosure, and the other with the coil floating.

One of the things I have noticed is that the length of the leads from the coil have a significant effect on frequency responses.  In some plots, you can see a small spike at about 23MHz.  This is due to the wire from the end of the coax to the beginning of the coil winding tight against the core.  The design of the enclosure specifically aims to minimise the unwound wire ends of the coil wire, but not on the prototype 3D printed enclosure.  I don't have a socket fitted to the enclosure as I would with a final version.

What to do

There are a few things I can do to improve the performance, by altering the cavity resonance:
1.  Not use an enclosure.  Increases the risk of e-field noise polluting conducted current emissions.  Not a good option.
2.  Increase the cavity resonance frequency by reducing the enclosed volume of the enclosed cavity.
3.  Dampen the resonance by filling the internal volume with very small ferrite beads.  This may be detrimental to current probe sensitivity.
4.  Try adding radial cuts in the aluminium shielding to impede the surface currents generated by resonance.  The effects of this are not yet known.

It is obvious that the enclosure has a significant effect on the current probe performance.    I am going to order a bag of ferrite beads from China.  The lead time is about 2 to 5 months. 
In the mean time, I will try changing the shape and size of the enclosure, and anything else I think of.