DIY RF EMC Current Probe Set Design

[dazz1]

It's amazing what you can accomplish with a clear bench !
For a good while that Wavetek was looking to be your nemesis and a damn fine thing it's behind you so that fog can clear and allow you to get on with this stuff that obviously has been tucked away for a bit.

Carry on Daz and still can't believe it's a year since I dropped in while down your ways.

The really great thing about antique test gear is that it is fixable with readily available parts.   It was the first piece of equipment I have fixed with so many faults,  I lost count.
Call in when you are next down this way. 

[jonpaul]

Hello again Dazz1, bravo for the progress.

A few thoughts

1/ use copper foil not aluminium, better shield, easily available and solderable

3M and others make shielding tapes in various widths and thickness.

2/ change winding method  to reduce self capacitance to core and turn to turn.

3/ The distributed term can be tested easily, just wind two or three sections, not 10..20 as Pearson.

4/ Lowest abberation widest bandwidth with coil's natural Zo, not 50 Ohms. Try Optimization of Zo say 25..200 Ohms first.

Afterwards use a resistive pad to match the result to desired 50 Ohm BNC/ cable.

Bon courage

Jon

[dazz1]

Hello again Dazz1, bravo for the progress.

A few thoughts

1/ use copper foil not aluminium, better shield, easily available and solderable

3M and others make shielding tapes in various widths and thickness.

All valid points but Cu foil is not available here at any sort of reasonable price.

2/ change winding method  to reduce self capacitance to core and turn to turn.

What you are looking at is a 9 turn coil.  Early testing indicates that self capacitance is not a problem.
Close coupling to the inner shield has proven to be essential to damp oscillations and improve bandwidth.

3/ The distributed term can be tested easily, just wind two or three sections, not 10..20 as Pearson.

The mechanics of connecting resistors to a coil and to a ground would be time consuming.

4/ Lowest abberation widest bandwidth with coil's natural Zo, not 50 Ohms. Try Optimization of Zo say 25..200 Ohms first.

I haven't done any impedance/matching testing yet.  It's on my list.

Afterwards use a resistive pad to match the result to desired 50 Ohm BNC/ cable.

yes.

Bon courage

Jon

[Hydron]

If you need Cu foil tape in a hurry, try looking for "snail tape" in a gardening store. Will not have conductive adhesive, but you can still tack solder it to adjacent pieces. Also, it looks like trademe does have local sellers offering it affordably (including with conductive adhesive)?

[dazz1]

If you need Cu foil tape in a hurry, try looking for "snail tape" in a gardening store. Will not have conductive adhesive, but you can still tack solder it to adjacent pieces. Also, it looks like trademe does have local sellers offering it affordably (including with conductive adhesive)?

Hi
Thanks for the suggestions.  I hadn't heard of "snail tape".  I hate gardening.
I suspect that gardening store and Trademe copper tape is most likely copper plated aluminium based on the very low price compare to 3M real Cu tape.
There are quite a few sellers on Trademe that buy stuff from Aliexpress and then sell it locally.  I buy a lot of stuff directly from Aliexpress.
A lot of products with copper wire sold on Aliexpress are made with copper clad aluminium.  It looks exactly like copper wire, but isn't.

The final machined metal version of this current sensor is going to be made from aluminium.  Even if I could buy a big enough piece of copper, the price would be many $
So kitchen foil is a reasonable and practical (ie.  low cost) proxy for solid aluminium.

[dazz1]

Hi
The next prototype No.8 now includes a BNC connector.   I will not be using a BNC connector in the final metal version because of its size. It would require a larger diameter piece of bar to fit within an eccentric shape than an SMA connector. 

In this version, I also have shells of different diameter to investigate the resonance damping effects of the inner shield.  If the inner shield diameter is too great, there is a lot of resonance.  If the diameter is too small, maybe I am getting too much damping.  That will degrade sensor sensitivity.

[wilhe_jo]

Hi!

You're overcomplicating things...

Incidentally, I made some EMC probes myself just lately.
I do EMC-consulting as a business, and having really cheap probes allows me to have plenty of probes to leave several test-setups unchanged.

If you read the standards, you'll want <1Ohms of Insertion Impedance. So 7 Turns are slightly too little; 8 is fine.
Just use a simple coax. This gives you a nice electrostatic shield, is cheap and just works well enough for 150k-30MHz

In the attached photo, one uses a Epcos core, the other one sold by Würth.
The 6 winding one was a test to see if the simple impedance transformation formula works out for RF (and this arrangement) as well.

I plan to have this semi-open-source... ie. the 3d-printing files+plans for one make of core and a generator for arbitrary core dimensions (this is already finished in OpenScad) sold for cheap,
Maybe I'll also get a cheap kit out with all parts included.
I'm fully sure if there would be a "market" for this, but this may be doable for 100€ including international shipping... so I tend to give it a try in the near future.

I still have to do some verification work.
This will be mainly testing if the 3d prints hold up to common miss-use in a lab  (ie. does the hole for the BNC survive or do I need to increase the wall thickness).
Another thing to verify is the shielding effectiveness... I hope to get this done  the next week.

For the cal-jig, I opted to some PCB-"cage", as you can see in the other attachment (that's just a rendering for my documentation...).

The cal-jig is fine up to about 100MHz and is useful to characterize random cores as long as they fit into the jig
My probe is fine up to 150MHz. So that's plenty good for conducted EMC testing.


For your bi-conical antenna... I'd replicate the one in MIL STD-461.. there are plans in the standard and you get some reference antenna factors.

[dazz1]

Hi!

You're overcomplicating things...

Yes I am.    Over-engineering is fun 

Incidentally, I made some EMC probes myself just lately.
I do EMC-consulting as a business, and having really cheap probes allows me to have plenty of probes to leave several test-setups unchanged.

If you read the standards, you'll want <1Ohms of Insertion Impedance. So 7 Turns are slightly too little; 8 is fine.
Just use a simple coax. This gives you a nice electrostatic shield, is cheap and just works well enough for 150k-30MHz

I did 9 turns but I haven't tested impedance yet. 
OK which standards?  I haven't seen any for the current sensors. 
I started with the hope of to get something that would work 100k to +30MHz.  I never expected to get to where I have. 

In the attached photo, one uses a Epcos core, the other one sold by Würth.

Core model numbers?

The 6 winding one was a test to see if the simple impedance transformation formula works out for RF (and this arrangement) as well.

And did it work out?

I plan to have this semi-open-source... ie. the 3d-printing files+plans for one make of core and a generator for arbitrary core dimensions (this is already finished in OpenScad) sold for cheap,

Nice.

Maybe I'll also get a cheap kit out with all parts included.
I'm fully sure if there would be a "market" for this, but this may be doable for 100€ including international shipping... so I tend to give it a try in the near future.
I still have to do some verification work.
This will be mainly testing if the 3d prints hold up to common miss-use in a lab  (ie. does the hole for the BNC survive or do I need to increase the wall thickness).

Do you think that the two legged connector holder will be strong enough? 
I have found that even a really short length of exposed end of the winding, not wrapped close around the core, is a problem.

Another thing to verify is the shielding effectiveness... I hope to get this done  the next week.

Please share the results.  If coax shielding works well, Pearson is out of business.  
Do you just ground the coax shield at one end to avoid a shorted turn? 
Do you see any resonance?

For the cal-jig, I opted to some PCB-"cage", as you can see in the other attachment (that's just a rendering for my documentation...).

I like that.   
My uncharacteristically simple fixture is reasonably good I think but I will record a plot.

The cal-jig is fine up to about 100MHz and is useful to characterize random cores as long as they fit into the jig
My probe is fine up to 150MHz. So that's plenty good for conducted EMC testing.

Have you done a frequency response plot?

For your bi-conical antenna... I'd replicate the one in MIL STD-461.. there are plans in the standard and you get some reference antenna factors.

I have based the dimensions on the MIL-STD but I have changed the mechanical construction to simplify manufacture and to make it snap assembly.  I will make it to pack flat.    The top of the antenna arms are curved so the ends enter the top junction at 90 degrees.  It will be a lot easier to bend the arms and drill the junction sockets at 90 degrees.    The curved arms will be closer to the ideal shape for a bi-conical antenna, but I think the most significant effect will be people asking me why they are curved.

The antenna will be cheap and easy to make, so I may make several variants, depending on how well they work.
I am planning to make two identical antenna for a start, because I will then be able to calibrate them.

I haven't yet figured out what I am going to do for a balun.  There are a few options.

[wilhe_jo]

Actually, the probe is defined in MIL-STD-462 and CISPR 16-1-2.

In the end, using this probe is an established test procedure in the EMC world.
So you get considerations for the measurement uncertainties and even construction plans right from the standards

I was a little wrong... it was Kemet ESD-R-57D-1 and Würth 74270097.
The Würth one is especially interesting because of its price.

The lower you get with your number of turns, the higher the resonance frequency gets.
So 8 turns is what you want  to get well below 1 Ohms of insertion impedance.
And yes, the simple impedance transformation formula is good enough even in this crude coupling situation.

So far, my 3d print held up well.
However, I think, I'll add maybe one more perimeter to get some more stiffness when printing with PETG... PLA is plenty sturdy already.

With regard to measurement results... they are in spec, but I haven't finished the documentation (I will have a presentation on a EMC symposium where this will get some closer discussion).
A resonance is there, but that's somewhere around 180-200MHz IIRC.

Using a coax is plenty good... to get the best bandwidth, just connect the shield on both ends and cut it at half the length (ie, you have half of the length from both sides).
If I'm not totally wrong with my estimates, there's a resonance reflecting 1/4 lambda for the shield... so yes, doing this in this cheap/economical way has some drawbacks.

There are some more things in the CISPR standard to consider: Transfer Impedance, shielding effectiveness against external electric fields and induction from external magnetic fields.

So far, I'm happy with the results and I hope to get the rest of the measurements done in the next weeks.
As I said, I have kind of a deadline because I'll use some of this stuff in a presentation about DIY stuff in an EMC lab

73

[dazz1]

Hi
Today I did some testing with an RF bridge on the Spec An.
The aim was to test the fixture and the effects of a coil in the fixture.


I started off testing the Suhner 50ohm BNC load.  This also showed the performance of my DIY RF balanced bridge.   The one I use is shown here:  https://www.epanorama.net/blog/2017/10/06/rf-bridge-for-antenna-measurements/.  They are available from multiple suppliers.  If fitted mine into a cast aluminium box that is a really good fit.  Doing this creates problems with internal resonance that mangles the performance.  I fixed this by gluing lots of small ferrite beads onto the interior surfaces of the enclosure. 

The first image (1) below shows the test setup laid out on the bench.
The RF bridge was connected to the test fixture and not to the coil for all tests.
All tests were done from 0 to 500MHz
The Spec An was setup to display reflected power and VSWR.  100% reflection from the load end gives a trace at 0dBm.  When the load absorbs nearly all of the power, the reflected signal is very low. Much less than )dBm.

The first trace below (2) shows how well the Suhner BNC 50ohm load works.   This was connected directly to the DUT port of the RF Bridge. This is at least 40 years old so it was made long before Aliexpress, Bangood and others were around.

For comparison, the next trace  (3) shows the response of the cheap SMA Aliexpress 50ohm load.  This was connected directly to the DUT port of the RF Bridge.  The response of the cheap Aliexpress load is arguably better than the Suhmer.  The Suhmer had the disadvantage of needing a cheap BNC to SMA adapter to connect to the bridge.  This could explain the results.  The test data shows that both test loads work well.

The next trace (4) shows the effect of adding the coil fixture between the bridge and the 50 ohm  load.  No coil was fitted to the fixture.   In this configuration, all of the energy from the tracking generator should pass through the fixture and be fully absorbed by the 50ohm load.  Ideally, no RF should be reflected back from the fixture.  The first tests showed the 50ohm loads reflected almost no RF.  For practical purposes, all of the reflection is from the test fixture.    The reflection trace shows the RF test fixture has exceptionally ordinary performance.  By the time it gets to 250MHz, almost all of the RF is reflected. Less than 1dB makes it through to the test load.    So when I am looking at a coil response trace at 200MHz, only about 2.5dBm is the coil under test.

The next trace (5) shows the effect of adding a coil to the fixture.  The coax connected to the coil is terminated with a 50 ohm load.  This looks at the effect of wrapping a coil around a wire carrying current.  Any voltage induced in the coil is going to be absorbed by the 50ohm load connected to the coil via the coax. 

To test the effects of the coil and connected load, I removed the test load from the coax.  Any voltage induced in the coil would be reflected back by the open circuit at the end the coax.
Spurs can be seen (6) in the trace where the resonated back to the coil.  The resonance in the coil and coax was induced back to the wire in the test fixture.  This shows that what is connected to the coil does affect the current passing through the middle of the sensor coil, but not by much.   If there was no resonance, it would be less obvious that there is any difference.


So I really should try and make a better test fixture.  By the time this one reaches 100MHz, the results are unreliable.    It works well enough to tell the difference between two coils, but not the absolute values.
What I would do is adapt a TEM cell design like this one for coil testing:

    

TEM cells are not suitable for coil testing because the central conductor is a sheet of metal that is far too big to fit through a coil.   The sheet metal top and bottom connected to the coax shield are shaped to match the cell to the coax and reduce reflection.  Achieving good matching includes the shape of the central conductor.  A TEM cell based coil tester does not need to be as large as the one shown in the link.
Replacing the central sheet conductor of the cell with a metal wire or tube should retain a reasonable match to the coax and to the load.  It won't be perfect, but it would be better than my fixture or the Pearson version.

[wilhe_jo]

Well, IMHO a current-probe-cal jig can never get reasonably "well matched" (that's the wording from the standard) to 50 ohms...

That's just physics... just get some of these coaxial line characteristic impedance calculators and enter some numbers...
You'll get useless dimensions for an air-dielectricum at 50ohms.
Even the jigs shown in the standard will never ever get close to a match.

However, you can certainly some signal source and measure the voltage/power into some defined load (ie. do something like an S21 measurement).
During this, you can investigate the power/voltage at the output of your probe.
Now you know the current through the center contact at each frequency, and you know the output of the probe.
This gives you the transfer impedance.
That's the suggested method in the standard, btw.

I see similar values like you. Up to around 100MHz, you get no problems.
My intention is not to go way beyond that. So I don't really care

The nice thing with this is, you can terminate the jig with 50ohms and do a S21 measurement between the output of the probe and the jig.
This gives you everything (considering the jig does not influence the match too much).

However, I have everything to do a "correct" calibration.
So at one point, I'll take all my current probes and do the correct setup with some power meters and a sig-gen (+ maybe some amplifier - just for fun .

73

[dazz1]

Hi
I made my test jig before I had seen the Pearson version and when I thought I'd be struggling to get to 30MHz.  I recognise that the flaw with both versions is  the return signal has to make 4 right angle turns.  At every turn there will be reflection.    Also, the two abrupt transitions from coax to sheet metal are also reflective.   So the theory of doing a TEM cell shaped like jig is that the transitions and corners are softened.

Using a coax impedance calculator, I came up with some achievable dimensions for a coil tester.   Making the central conductor with a larger diameter has a significant effect  on impedance. 
Approximating a coax with a TEM cell shaped object should provide a better match above 100MHz.    TEM cells are good to a few GHz.  Maybe I can get to a couple of hundred MHz.

Do I think your/ the standards method of measuring S parameters will work?  Yes
Do I have the test gear to measure S parms? Yes
Do I need to measure above 100MHz? No
Do I have the materials and tooling to knock up a TEM-shaped test rig in less time than measuring the S-parms? Yes
Will anyone care ? I doubt it!

[wilhe_jo]

Well, I came up with similar numbers...

The problem is, I need around 70mm outer diameter and I have a 25mm inner bore.
Additionally, adding the Probe will change impedance drastically....
After having some beers about this problem, I just tabled (US meaning!) it.

I guess it is very reasonable to do something similar to CDN calibration.
I.e.: calibrate your "forward power" to get some reasonable output at the termination and measure the probes output in a second step.
My hope is to have higher dynamic range using power detectors (or even the receiver) than with my VNA... this would give me better values for the lower frequency ranges.

Even though you get some different impedance along the line, the current (ie. the integral of the current density along the conductor area) will be the same - regardless of that impedance jumps.
Hence, we're golden

A open TEM cell is limited by its length... mostly.... BTDT

There are nice plans out there.
A 300MHz open TEM cell is nice, but my 750mm GTEM is nicer

73

[dazz1]

Well, I came up with similar numbers...

The problem is, I need around 70mm outer diameter and I have a 25mm inner bore.
Additionally, adding the Probe will change impedance drastically....
After having some beers about this problem, I just tabled (US meaning!) it.

I am going to give it a try and see what results I get. 

I guess it is very reasonable to do something similar to CDN calibration.
I.e.: calibrate your "forward power" to get some reasonable output at the termination and measure the probes output in a second step.
My hope is to have higher dynamic range using power detectors (or even the receiver) than with my VNA... this would give me better values for the lower frequency ranges.

Even though you get some different impedance along the line, the current (ie. the integral of the current density along the conductor area) will be the same - regardless of that impedance jumps.
Hence, we're golden

Agreed, unless there is a C storing energy in the middle.
It would be a lot easier if the current is reasonably flat over the working frequency. 
I don't have a VNA so I can only measure s-parms magnitude, not vector values.

A open TEM cell is limited by its length... mostly.... BTDT

There are nice plans out there.
A 300MHz open TEM cell is nice, but my 750mm GTEM is nicer

73

My collection of test equipment is very old (except for the Siglent spec-an) and modest.

[dazz1]

Hi
I am made and tested my new coil current test jig.  It looks similar to a TEM cell, but it isn't.  It is intended to approximate a short length of expanded coax cable. 
The outer conductor is two pieces of sheet aluminium that provide a relatively smooth path from one end to the other. 
In order to achieve a reasonably matched 50ohm impedance, the centre conductor, consisting of a bar of machined aluminium, tapers at both ends. 
The bar diameter is constrained by the diameter of the coil that it passes through.  As a result, the bar diameter is not sufficient to achieve 50ohm impedance.

The bar is dimpled at both ends.  Bronze brazing wire was machined to fit into the BNC terminal and soldered in place.  This left a 3mm short spike of brazing wire sticking out the back of the connector.  The bronze spikes fit into the 1mm deep dimples. The bar is held in place by spring compression of the sheet metal.    To remove/insert the bar, I just squeeze the sheet metal so the connectors spread apart.  The bar drops out.  Very simple to make and use.

In the next few posts, I will attach test results.
If it doesn't work, it will make a great coffee table conversation starter.

[dazz1]

Hi
Attached is the datasheet for the Pearson calibration fixture.  I can't find a listed price for this item, so I guess the price probably comes with a health warning.
Consult your doctor first.  You might get a heart attack if you see the price. 

[dazz1]

Hi
The first two attachments are of the new and old test jigs out to 1GHz.
SWR is marked at 4 different frequencies.
The new TEM cell like jig is >3.5db better return loss than the first U shaped jig. 
The SWR is smooth across the frequency range with no sign of any resonance.

The Pearson jig claims a frequency range of 10 kHz to 400 MHz with VSWR < 3. 
I am measuring VSWR 2.3 @ 800MHz with the new jig.  At 800MHz, the old jig I am measuring VSWR = 14.

So the new TEM-like jig is significantly better than my 1st attempt and at least as good as the Pearson jig.   

[dazz1]

This is a repeat of the test in the previous post but out to a frequency of 500MHz.
Again the new test jig is significantly better than the old version. 

Both tests are of the bare test jigs alone.  No coil is fitted, so these tests are looking at the performance of the test jigs only.
My DIY RF balanced bridge is used to measure the VSWR.  The test jigs are terminated with 50ohm loads.  Previous tests show these perform well with good VSWR measurements.   

Comparison of the VSWR plots for the Pearson and the new TEM-like test jig indicate my jig is better.

[wilhe_jo]

Tekbox has similar cal-fixtures for IMHO a reasonable (read: reasonable if you buy this for a business) price.
The VSWR suggests, it will also have a resonance around 120MHz... so not "better" or worse than our tries.

Anyways, I'm not sure if the VSWR really matters too much.
The probe will definitely disturb the dielectric (less air - more conductive stuff).


Do you have any programmable signal source?
You could easily sweep the frequency and get the reading with your SA for both, the "output" of the fixture and the output of the probe.
In the end, a SA is "just" a super selective power meter on steroids.

Your SA has 50Ohme input.
That could perfectly serve as the load for getting a reference reading to obtain the current while you terminate the probe with some 50Ohms terminator.

When measuring the probes output, you could just move the terminator to the output of the fixture, and you're done.

BTW: I'm also in the process of making a measurement automation software... I already have a similar measurement - so that would be easy to implement.

73

[dazz1]

This is a different test.  The purpose was to directly measure the power that passes through the test jig across the frequency range.
So rather than looking at reflected power (VSWR), this test measures forward power directly at the load side of the test jig.

The tracking generator was set with the output going to one side of the test jig.
On the other end of the test jig, the 50ohm load was replaced with a 10dB 50 ohm attenuator. 
In effect this is like putting a volt meter on the load to measure how much power gets through the test jig. 
With this setup, a perfectly flat output would show as -10dBm across the frequency range.

The plots attached are across a 1GHz frequency range.  Even the bad old test jig performs reasonably well.
EDITED: Correct plots attached.

[dazz1]

Tekbox has similar cal-fixtures for IMHO a reasonable (read: reasonable if you buy this for a business) price.
The VSWR suggests, it will also have a resonance around 120MHz... so not "better" or worse than our tries.

This is a measurement of the test jig only.  No coil is fitted to the jig.  It will have no effect on the resonance of the coil.

Anyways, I'm not sure if the VSWR really matters too much.

I think it does matter in that it is a measure of how much current actually gets through the test jig. 
See my latest post with the 50ohm load replaced with a 10dB attenuator.

The probe will definitely disturb the dielectric (less air - more conductive stuff).


Do you have any programmable signal source?
You could easily sweep the frequency and get the reading with your SA for both, the "output" of the fixture and the output of the probe.
In the end, a SA is "just" a super selective power meter on steroids.

I think have already done this test, but given that VSWR measures reflected power, I could have just done this test with a calculator.

Your SA has 50Ohme input.
That could perfectly serve as the load for getting a reference reading to obtain the current while you terminate the probe with some 50Ohms terminator.

When measuring the probes output, you could just move the terminator to the output of the fixture, and you're done.

BTW: I'm also in the process of making a measurement automation software... I already have a similar measurement - so that would be easy to implement.

That would be really useful, especially if it could be connected to a stepper motor driving a turntable.

73

[dazz1]

Attached are the results of the same tests of VSWR.

The marker table shows the measurements at specific frequencies. The measurements show that the TEM-like test jig is a significant improvement over the original.  It will easily measure out to 500MHz with high confidence.  There are no resonance or other bad characteristics.   When compared to the Pearson test jig data sheet, the TEM-like jig is at least as good, and probably better.  This is the Mk1 Mod0 version so I am sure the performance can be improved but I am not going to do that.  One way to improve on this would be to use tube and hollow cones to actually make a full coaxial conductor pair.  I think this would be unnecessary.    The test show the TEM-like test jig is a reasonable approximation of an expanded coax conductor. 

The TEM-like test jig is simple and easy to make.   This one can accommodate a coil with a max external diameter of 65mm.  The centre conductor is 18mm diameter.  It would be easy to make different diameter versions for smaller/larger coils.

So in conclusion, the TEM-like test jig is a significant improvement over the first one I made.    The frequency range extends far beyond what I need for testing current sensing coils. 
This line of experimentation has been a success.

[wilhe_jo]

I think my comment re the VSWR needs some explanation...

Of course, it affects the power into the cal-system, but when you can measure the voltage at the load ( ie just use the SA as a 50 ohm load), it does not matter.

However, if you put your probe around the septum of your stripline/tem cell/coax will see significant change as it modifies the dielectric.
Having that said - as long as this mismatch is short compared to your min wavelength it shouldn't matter too much.

As I said, I tried to get some nice values for my coax adapter but the closest I could get was abt 100ohms. So I completely ignored that...
The only reason I see why a reasonably good VSWR would be nice ist that you could feed the jig from an amplifier.

Some 10W into the load would allow for lower frequencies to be measured. However, I tend not to bother about everything <150kHz.
Those commercial probes are often specified down to 100Hz. So every dB of input power helps lifting the output signal above the noise floor.
But even here I see no problem. My PA can handle mismatch and a 6dB attn would solve this issue anyways (this commonly done for conducted immunity testing).

Re my test software: I'm currently designing a manipulator for my GTEM.
That's basically turning a rod with some motor by 120degree. So turntable control should be there soon as well.

A cheap hacked combined with a stepper+some platform makes a nice antenna pattern measurement tool.
I did this in the past in a hacky way to quickly get some plots.
That will be handy for a current project, so I'll definitely add support for this

I hope everything will be "ready" and polished in september....

74

[dazz1]

Hi
All of what you say makes perfect sense.
I don't think a 50ohm test jig is practical given the physics of the problem.   I agree that VSWR isn't important.
I was far more interested in building a jig that degraded gracefully as frequency increased.  Specifically I didn't want any resonance.

With a relatively flat VSWR curve, it won't be so necessary to account for the characteristics of the test jig when measuring the coil.  Knowing that the current reaching the 50ohm termination is almost constant with frequency is useful.  Of course inserting a coil will mess that up.   Knowing that direct measurement of the termination current will give high quality data is another benefit.   Based on the limited data for the Pearson version, I think the measured data shows the TEM-like jig has superior performance. 

Of course the most important thing is the looks.  To a non-rf engineer, the purpose and how-it-works will be a complete mystery.    That's the thing I like about RF.   With some low tech tools and materials, I can make something with high tech performance.

[dazz1]

I just realised, I forgot to post the termination power measurements.
These were made by replacing the 50ohm termination load with a 10dB attenuator.  This directly measures the current that has passed through the test jig.
The plots show power to 1GHz and 500MHz, for both the old jig and the new TEM-like jig.

The new jig has a very flat curve, <1dBm to 500MHz.  It is still usable out to 1GHz with no resonance or other issues.

The attached image shows the test setup with a10dB attenuator replacing the 50ohm load.

The other attached images shows the new and the old test jigs.  The central conductor of the new jig is held between two spikes of bronze soldered into the BNC connectors.  The sheet metal provides spring force that ensures good contact with the aluminium central conductor.