Howto Calibrate DIY EMC pre-compliance antenna??

[dazz1]

Hi
I have a need for pre-compliance testing of a product I am developing.  I plan to purchase a Siglent 3000 series Spec An.    Of course by itself, it is useless for EMC testing without the sensors and accessories needed to measure stuff.  One of those thing I will need is a decent bi-conic antenna.    The commercial ones seem to be priced in gold.

I can design and build an antenna but the problem is then to calibrate it to a reasonable degree of accuracy for pre-compliance testing.  One procedure is to use a calibrated antenna as a signal source to test and calibrate the antenna under test (AUT).    But that requires one of those golden calibrated antennas.

I propose to overcome the need for a golden calibrated antenna by building and testing two identical, but unknown antennae.   The idea is simple (which probably means others have already thought of it and done it). 
Step 1.  Design/Build two identical antenna.
Step 2.  Connect cables / attenuators etc to the Spec An. 
Step 3.  Normalise the test rig by doing a sweep with the antennae replaced with a short coax link.
Step 4.  Insert antennae and do the tests.
Step 5.  Halve the test results.

This is illustrated in the attached drawing.

The attenuators would provide wide band 50ohm closely matched to the Spec An impedance to miminise resonance in the test rig.
The antennae would need to be mounted far enough apart to be in the far field.
The test location would need to be rf quiet and without reflections.
The null link between the attenuaters would need to be as short as possible (an adapter).
Both legs of the test rig (Tx and Rx sides) should be identical for all sorts of reasons.

So the test results would then simply need to be halved to get calibrated performance. 
If each antenna had 5dB gain, the test curve would show 10dB.
If each antenna had a 20db loss, the test curve would show 40dB.
So by halving the figures measured by the Spec An, a calibration curve would be  produced.

The method can be checked at a single frequency by using two identical dipoles whose performance can be mathematically calculated.  A ring-in for a golden calibrated antenna.

So the outcome of testing two identical but unknown antennae would be a known calibration curve.  In addition, there would be a spare antenna with its own calibration curve.   

Some might think that building 2 antenna when you only need one is a waste.  Building 2 antenna is not much more work than building one when time taken to design, buy parts and test are accounted for.  The 2nd antenna could then be sold or kept as a spare.

I would be interested to know if anyone else has tried this and if so, what were the results?



Dazz

[Jay_Diddy_B]

Dazz,

Welcome to the forum !!

If you go to this website:

http://everyspec.com/MIL-STD/MIL-STD-0300-0499/MIL-STD-461A_8679/

You can download MIL-STD-461A, make sure you get the A version. starting on page 25 is a detailed set of drawings for making the bi-conical antenna. If you stick to the dimensions it should have the antenna factors published in the document.

Here is one of the drawings:



There are instructions in the MIL-STD-461 for measuring antennas, using two identical antenna.

I built one of these antenna and found that it was 'next to useless' without a proper anechoic chamber. There was too much background noise.

I found an open TEM cell to be really useful and works well.

Regards,
Jay_Diddy_B

[dazz1]

Hi
Thanks for the reference.  I live 30 minutes from a rural site located away from the city RF noise in a  deep valley. There is a cell site about 2km away.  Apart from that, it should be relatively RF quiet.  A biconic should work.

I hadn't considered a TEM cell.  It would be easier and cheaper to build than a biconic.  I will investigate that option further.
Do you have plans or design method?

Dazz

[dazz1]

Hi
I have found a standard that includes a design  SAE J1113-25 for automotive EMC testing.
I found a paper that analyses the performance of that design in detail. http://ieeexplore.ieee.org/document/950717/ That should include a lot of detail useful for any DYIers. 

TEMS seem to be used more for immunity testing rather than emission testing.

[Jay_Diddy_B]

I hadn't considered a TEM cell.  It would be easier and cheaper to build than a biconic.  I will investigate that option further.
Do you have plans or design method?

Dazz

Dazz,

Try:

https://pdfs.semanticscholar.org/7b3c/c56771978fb7dab27ee4ccc31d67f7d3167f.pdf

If you google images 'open tem cell' you can get some ideas.

Regards,
Jay_Diddy_B

[EE-digger]

Dazz,  I'm assuming you know what band you need but I have to say I hadn't pulled the biconic out to use in a number of years.  Most problems with low to moderate speed clocks and switching regulators have tended to be in the 100-200MHz,  400-500MHz or 900MHz and a bit higher.  The switching regulators are all running in the 250kHz to 2MHz range but harmonics, especially those from poor layout on switching nodes, tend to show up exceeding limits in the first band I mentioned with various display clock harmonics in the second band.

So, for the most part, I use a 200MHz to 1.3GHz log periodic of fairly compact size (it's an older Emco, don't have the model handy).  For all else, I just wait until we do a pre-scan or final tests at a local test lab.

[hendorog]

Hi Dazz,
I think you are neglecting the path loss between the antennas in your idea?  There will be a chunk of free space path loss between the two antennas compared with the coax adapter which you normalised against.

You could make a couple of identical 'reference' antennas though and then compare other antennas to them. The measurement results won't be absolute as you don't know the precise performance of the reference antennas. However you might be able to get a rough idea of their performance if the reference antennas were a design which you can model or simply find on the internet.

[tautech]

Dazz
These could be well worth looking at for a cheap and characterized EMC probe set:
https://www.eevblog.com/forum/testgear/new-affordable-emc-probe-set-at-a-bargain-price-from-ariel-rocholl's-lab/

[dazz1]

Hi
At this stage I am trying to map out the gear I need to do budget pre-compliance.  My local EMC test facility is a few hundred kilometers away so even a pre-scan is a major exercise in time and cost.  It makes economic sense for me to spend money on test equipment rather than waste it on failed compliance tests. 

I has been 20 years since I last did EMC testing but I was designing/testing buildings, not equipment.  I am out of date with everything and my current test gear is old and limited.  I don't have any EMC instruments.

I'd really like to get the Siglent 3032 but I probably only actually need the 3021 with the EMI add-on.
I can make some H and E probes easily enough.
My products use lots CE compliant (MeanWell) 90W 5V SMPS.  I will need to look at load side noise getting onto the SMPS input side   so I am sure I will need a LISN.
The TekBox CISPR 16 LISN unit looks like a good candidate .
I like the idea of a log period antenna simply because it will be a lot easier to store that a TEM cell or biconical.    I would really like the antennas to be calibrated so I can get reasonably close to the test lab results before it gets to the test lab.
I can make up some RF current probes.
Transient limiter for the Spec An
RF pre-amp.
Cables/adapters
chattering relay (RF noise generator)

That is my shopping/make-it list so far. 

Not sure if I need an ESD tester.  My product is in an all plastic custom moulded enclosure.   The metal chassis is not exposed.

I have found recommended lists here: http://www.interferencetechnology.com/assembling-low-cost-emi-troubleshooting-kit-part-1-radiated
-emissions/
and here: https://interferencetechnology.com/assembling-low-cost-emi-troubleshooting-kit-part-2-immunity/
I think I have the essentials on my list.

I haven't yet determined what testing I must do.  My product is a collection of CE compliant units that are wired together.  The exception is a custom interface board with no clocks or other noise generators and a high powered batter charger. The battery charger unit is a strong candidate for RF noise because of the high currents and high ripple factor (60%@20A).

[dazz1]

Hi Dazz,
I think you are neglecting the path loss between the antennas in your idea?  There will be a chunk of free space path loss between the two antennas compared with the coax adapter which you normalised against.

You could make a couple of identical 'reference' antennas though and then compare other antennas to them. The measurement results won't be absolute as you don't know the precise performance of the reference antennas. However you might be able to get a rough idea of their performance if the reference antennas were a design which you can model or simply find on the internet.

I thought of that. If the antennas are far enough apart to be in the far field, the free space loss can be calculated.  Far from perfect.  A pair of accurately made dipoles operated at their design frequency can be used to confirm the loss.  There is a lot of data on dipole performance.  Somewhat closer to perfection.   More data can be obtained by pairing a dipole (at the design freq) with the AUT.     No where near as good as a proper lab test but cheap and better than no test.

[dazz1]

Hi
Google found the Antenna test procedures here : www.diamondeng.net/library/AntennaMeasurement.pdf
Lots of info but a quick read indicates all methods described use a golden calibrated antenna.  A detailed read may reveal more.

Dazz

[hendorog]

Hi Dazz,
I think you are neglecting the path loss between the antennas in your idea?  There will be a chunk of free space path loss between the two antennas compared with the coax adapter which you normalised against.

You could make a couple of identical 'reference' antennas though and then compare other antennas to them. The measurement results won't be absolute as you don't know the precise performance of the reference antennas. However you might be able to get a rough idea of their performance if the reference antennas were a design which you can model or simply find on the internet.

I thought of that. If the antennas are far enough apart to be in the far field, the free space loss can be calculated.  Far from perfect.  A pair of accurately made dipoles operated at their design frequency can be used to confirm the loss.  There is a lot of data on dipole performance.  Somewhat closer to perfection.   More data can be obtained by pairing a dipole (at the design freq) with the AUT.     No where near as good as a proper lab test but cheap and better than no test.

Sorry for the delay, I didn't notice the reply notification for some reason.

Yep that is exactly what I was implying for the DIY reference antennas. I haven't specifically tried it, but you could do quite a few tests on your DIY dipole references with the SA+TG to see how close it is to the theoretical build. For example you could rotate one around while keeping the other one static and work out what the pattern looked like. And you could measure the return loss to work out the resonant frequency compared to your goal.

For a dipole you would need a balun to get reliable measurements, which you could also build and and test.

[SWR]

I'd really like to get the Siglent 3032 but I probably only actually need the 3021 with the EMI add-on.
I can make some H and E probes easily enough.

Thats what I bought and it's a really fine instrument.
You can max it out to 3,2GHz and all the options pretty easily if you're into that.

Instead of messing with antennas i use common mode chokes like the ones described here:
https://www.edn.com/electronics-blogs/the-emc-blog/4429621/Using-current-probes-to-estimate-E-fields

It's a much simpler solution and in my experience it gets close enough for useful pre compliance if you chose a high frequency ferrite. I cut a long SMA cable and pull out the inner wire a distance from the cut end. Then I wrap them round the ferrite and solder the ends together. The final touch are two 3D printed shells with grooves that keep the spacing of the coil turns in place. It's possible to get >50ohm up beyond 1GHz.

[dazz1]

Hi
Someone somewhere had to start with an uncalibrated antenna before they ended up with a calibrated version.  There are probably papers published  in the first half of last century by very smart researchers detailing the methods.

[Edit:  It turns out that someone has done that .  The Mil-Std461A referenced by Jay_Diddy_B includes a 2-antenna calibration procedure.]

Unfortunately the SSA3032 is the reason I own the 2.5litre version of my car rather than the 3.0litre engine.    The 3 litre engine is really nice but doesn't add any useable  capability for the greater cost.  The savings I can make in buying the 3021 I can spend on accessories that add significant capability.  I would get bigger bang for buck.  The choice is a personal one.  There is no right  or wrong answer that fits everyone.

The DIY chokes/current sensors are on my shopping list.

Dazz

[hendorog]

Unfortunately the SSA3032 is the reason I own the 2.5litre version of my car rather than the 3.0litre engine.    The 3 litre engine is really nice but doesn't add any useable  capability for the greater cost.  The savings I can make in buying the 3021 I can spend on accessories that add significant capability.  I would get bigger bang for buck.  The choice is a personal one.  There is no right  or wrong answer that fits everyone.

I'm pretty sure that the SSA3021 is hackable up to be an SSA3032 - if you want to. So you can have the best of both worlds there.

[dazz1]

I'm pretty sure that the SSA3021 is hackable up to be an SSA3032 - if you want to. So you can have the best of both worlds there.

I'd be very surprised if the SSA3021 had the hardware to go to the same max freq as the SSA3032 even if a software hack allowed it. If I was producing software for a SA, I'd use identical software for all models with separate configuration files.   That would reduce software production and maintenance costs.

An alternative option for me would be to buy the SSA3032 but sacrifice the software options.  EMC testing has been done for many years without the option of helper software apps. 

[hendorog]

I'm pretty sure that the SSA3021 is hackable up to be an SSA3032 - if you want to. So you can have the best of both worlds there.

I'd be very surprised if the SSA3021 had the hardware to go to the same max freq as the SSA3032 even if a software hack allowed it. If I was producing software for a SA, I'd use identical software for all models with separate configuration files.   That would reduce software production and maintenance costs.

An alternative option for me would be to buy the SSA3032 but sacrifice the software options.  EMC testing has been done for many years without the option of helper software apps.

I did a quick search and it does seem to be possible. It sounds like you are right on the software side as it's the config which is changed. I don't have one but the hackability of them does sound like a big selling point for a hobbyist.

Have a look here, where someone comments that their 3021 behaves like a 3032:
https://www.eevblog.com/forum/testgear/hack-of-sigllent-spectrum-analyzer-ssa3021x/msg1261248/#msg1261248

[SWR]

I'd be very surprised if the SSA3021 had the hardware to go to the same max freq as the SSA3032 even if a software hack allowed it. If I was producing software for a SA, I'd use identical software for all models with separate configuration files.   That would reduce software production and maintenance costs.

Well - I think you'll have to concider yourself very surprised then.
I converted my SSA3021 to 3,2GHz and enabled all the option while I was at it ... and it's all "Honky-dory" as they say.
You basically just have to rename a file so the analyzer can't find it.

[dazz1]

Well - I think you'll have to concider yourself very surprised then.
I converted my SSA3021 to 3,2GHz and enabled all the option while I was at it ... and it's all "Honky-dory" as they say.
You basically just have to rename a file so the analyzer can't find it.

It doesn't surprise me that the software allows it.  It doesn't surprise me that the hardware appears to work.  I would be surprised if a side by side comparison between a hacked 3021 and genuine 3032 showed the same performance.   That would require identical hardware on both.

[SWR]

It doesn't surprise me that the software allows it.  It doesn't surprise me that the hardware appears to work.  I would be surprised if a side by side comparison between a hacked 3021 and genuine 3032 showed the same performance.   That would require identical hardware on both.

They chose to go with one hardware and make the variation in software instead like many other manufacturers do.

It wouldn't make much sense to sell upgrades to a crippled hardware. They would face several challenges:
- Having to procuce separate PCBs and not gaining the benefit of low hardware variance.
- Dealing with unsatisfied customers that bought the upgrades without getting full performance.

It's all about doing the variants as late in the production process as possible.
A label and a software config is the optimum choise from a production point of view.

[dazz1]

Hi
I have been looking at the connection between the antenna and coax feeder.  I am aware that coax can be directly connected to the antenna cones.  This makes for a physically asymmetric antenna (including the antenna feed) which is likely to distort the antenna pattern.

I think a 1:1 balun would be a good idea.  This will allow the antenna feed to exit radially to avoid distorting the antenna pattern.   I found this research paper digitalcommons.calpoly.edu/cgi/viewcontent.cgi?article=1282&context=eesp.    The results indicate that the current transformer type balun has good performance.  The test frequency went to 1.5GHz.  I like measured data.

I don't know where to get suitable tiny ferrites.

Dazz

[RFDUK]

I went through the same search in the early 1990's setting up an EMC lab on a limited budget.
The solution I found is probably still a good answer to your question today.

In the early days of commercial EMC testing (as apposed to the established military EMC standards of the time), the FCC developed a 'reference dipole' for absolute field strength testing.
The guy responsible for the balun design was Willmar Roberts. The balun used simple lengths of 50 ohm cable.
'Roberts Balun' antennas cost a fortune in commercial sets like EMCO 3121, but can be manufactured for very little.

They are so simple that if built exactly to the drawings they are self calibrated to a very low uncertainty.

If you Google Roberts Balun reference dipoles hopefully the drawings are out there. They were initially published in an FCC document.

Your idea of building two antennas is a good one for sure. That's what I did with the Roberts Baluns, you need 4 baluns (and 4 sets of dipoles) to cover the range 30 MHz to 1 GHz.
By connecting two baluns for the same frequency range back to back (balanced connections), you can test the transmission performance integrity. The plots I still have somewhere looked great.

Testing 2 unknown antennas back to back poses the problem you mention of reflections. That isn't easily accommodated unfortunately.

Once you have a set of 'reference dipoles' you can calibrate other antennas you make, for example Biconical or Bilog etc. My homemade dipoles are still in use today 25 years later.

For higher frequencies very low cost PCB Log Periodic antennas are available from Kent at http://www.wa5vjb.com.
Don't be tempted by other products, Kent's antennas are the real deal.

Hope that's interesting, good luck, Martyn.

[dazz1]

Hi
Thanks for the info.  I will look that up.

The nice thing about simple dipoles is that both the theory and practice have been so thoroughly characterised.  There can't be much left unknown about them.

What I found really interesting about the baluns in the link was the tiny size and the very fine Cu wire used.  I would have thought that this would present a major impedance mismatch.

I was planning to use Cat6 twisted pair (100R impedance) 2x in parallel (50R).  That would require a much larger ferrite core.  That's one of the reasons I like measured data because it eliminates the BS. 

A Kent antenna is on my shopping list.

[dazz1]

Hi
I found this on the Roberts Balun.
http://dashfoundation.com/Archives/Roberts%20-%20FCC%20Antennas.pdf

Dazz

[RFDUK]

Hello Dazz, it's great to see that doc again thanks! That's the one I used. The photo shows what I did using plastic boxes and cable conduit parts. BNC sockets were glued into the conduit ends. The lowest frequency antenna rods extend to about 60MHz. I had extension wires with crock clips to get to 30 MHz.

I made 2 of each balun and the swept performances fitted the doc frequency response very well.

I made a bicon at the time too. A friend milled a center piece from Delrin and turned aluminium bosses for each end of the rods. I noticed the bicon diagram posted earlier had connecting parts in the rods. I just bent the aluminium rods and that worked fine. The balun was a wire transformer on low permeability ferrites. I found a low priced commercial 30MHz - 2GHz Bilog on ebay about 15 years ago so the homemade bicon is out of service now.

It's all ways useful to have proper reference dipoles available as a sanity check against broadband antennas, I wouldn't want to be without them.

Bicons need careful calibration below the natural resonance of about 70MHz for the larger ones. The VSWR is poor below the natural resonance and it's best to put a 6dB pad at the antenna end to remove K factor drift associated with mismatch to the cable.

Martyn.